WO2025145325A1 - Encoding method, decoding method, encoders, decoders and storage medium - Google Patents

Abstract

Provided in the embodiments of the present application are an encoding method, a decoding method, encoders, decoders and a storage medium. The decoding method comprises: analyzing a code stream, and determining a quantization coefficient of points in a three-dimensional mesh; performing dequantization on the quantization coefficient of the points in the three-dimensional mesh, and determining a dequantization coefficient of the points in the three-dimensional mesh; performing inverse transform on the dequantization coefficient of points in a first level to determine a displacement coefficient of the points in the first level; and on the basis of the displacement coefficient of the points in the first level, predicting a displacement coefficient of points in a second level, the points in the three-dimensional mesh belonging to a plurality of LODs, the first level and the second level being two adjacent levels amongst the plurality of LODs, and the first level being a higher level of the second level.

Description

Coding and decoding method, codec and storage medium Technical Field

The present application relates to the field of video coding and decoding technology, and in particular to a coding and decoding method, a codec and a storage medium.

Background Art

Video dynamic mesh coding (V-DMC) requires encoding the geometric information of points in the mesh. How to improve the coding efficiency of geometric information is a problem that needs to be solved.

Summary of the invention

The embodiments of the present application provide a coding method, a codec, and a storage medium to improve the coding efficiency of geometric information of points in a grid.

In a first aspect, a decoding method is provided, which is applied to a decoder, comprising: parsing a bit stream to determine quantization coefficients of points in a three-dimensional grid; dequantizing the quantization coefficients of the points in the three-dimensional grid to determine the dequantization coefficients of the points in the three-dimensional grid; inversely transforming the dequantization coefficients of the points in a first layer to determine the shift coefficients of the points in the first layer; predicting the shift coefficients of the points in a second layer based on the shift coefficients of the points in the first layer; wherein the points in the three-dimensional grid belong to multiple level of detail (LOD) layers, the first layer and the second layer are two adjacent layers in the multiple LOD layers, and the first layer is a layer above the second layer.

In a second aspect, a coding method is provided, which is applied to an encoder, comprising: performing detail level LOD division on points in a three-dimensional grid to determine multiple LOD layers, wherein the multiple LOD layers include adjacent first and second layers, and the first layer is an upper layer of the second layer; predicting the shift coefficients of the points in the second layer to determine the residual coefficients of the points in the second layer; and transforming the shift coefficients of the points in the first layer according to the residual coefficients of the points in the second layer.

According to a third aspect, a decoder is provided, comprising: a first determination unit configured to parse a bitstream and determine quantization coefficients of points in a three-dimensional grid; a second determination unit configured to dequantize the quantization coefficients of the points in the three-dimensional grid and determine the dequantization coefficients of the points in the three-dimensional grid; a third determination unit configured to inversely transform the dequantization coefficients of the points in the first layer and determine the shift coefficients of the points in the first layer; a prediction unit configured to predict the shift coefficients of the points in the second layer based on the shift coefficients of the points in the first layer; wherein the points in the three-dimensional grid belong to multiple LOD layers, the first layer and the second layer are two adjacent layers among the multiple LOD layers, and the first layer is the upper layer of the second layer.

According to a fourth aspect, a decoder is provided, comprising: a memory for storing a computer program; and a processor for executing the method described in the first aspect when running the computer program.

In a fifth aspect, an encoder is provided, comprising: a first determination unit, configured to perform LOD division on points in a three-dimensional grid, and determine multiple LOD layers, wherein the multiple LOD layers include adjacent first and second layers, and the first layer is an upper layer of the second layer; a second determination unit, configured to predict the shift coefficients of the points in the second layer, and determine the residual coefficients of the points in the second layer; and a transformation unit, configured to transform the shift coefficients of the points in the first layer according to the residual coefficients of the points in the second layer.

In a sixth aspect, an encoder is provided, comprising: a memory for storing a computer program; and a processor for executing the method described in the second aspect when running the computer program.

In a seventh aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed, the method as described in the first aspect or the method as described in the second aspect is implemented.

In an eighth aspect, a non-volatile computer-readable storage medium for storing a bit stream is provided, wherein the bit stream is generated by an encoding method using an encoder, or the bit stream is decoded by a decoding method using a decoder, wherein the decoding method is the method described in the first aspect, and the encoding method is the method described in the second aspect.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1A is a schematic diagram of a three-dimensional grid image.

FIG. 1B is a partial enlarged view of the three-dimensional grid image.

FIG. 2 is a schematic diagram of a connection method of a three-dimensional grid.

FIG. 3A is a schematic diagram of a three-dimensional grid image.

FIG. 3B is a schematic diagram of a grid data storage format.

FIG. 3C is a property diagram of a three-dimensional grid image.

Figure 4 is a diagram of the overall framework of grid coding.

FIG. 5A is a schematic diagram of a mesh preprocessing process.

FIG. 5B is a schematic diagram of a method of generating shift coefficients.

FIG. 6A is a schematic diagram of a quantization processing method for mesh geometric information.

FIG. 6B is another schematic diagram of a quantization processing method for mesh geometric information.

FIG. 7A is a schematic diagram showing an encoding method for the connection relationship of triangular facets.

FIG. 7B is a schematic diagram of a method for encoding geometric information.

FIG. 7C is a schematic diagram of a texture coordinate encoding method.

FIG. 8A is a schematic diagram showing the basic principle of vertex shift coefficients.

FIG. 8B is a schematic diagram of a mapping method of mapping a shift coefficient to a two-dimensional image.

FIG. 9 is a schematic diagram of an encoding method for inter-frame geometric information.

FIG. 10A is a schematic diagram of an intra-frame coding method.

FIG. 10B is a schematic diagram of an inter-frame coding method.

FIG. 11 is a schematic diagram of a basic grid division method.

FIG12 is a schematic diagram of the hierarchical structure of the basic grid.

FIG. 13 is a schematic diagram of a coefficient reorganization method of shifted coefficients.

FIG14 is a schematic diagram of a coding block structure in a two-dimensional image.

FIG15 is a schematic diagram showing the relationship among the basic grid, the shift coefficient and the reconstructed grid.

FIG. 16 is a diagram showing an example of the connection relationship of points in a three-dimensional grid.

FIG. 17 is a diagram showing an example of a method of determining a quantization parameter.

FIG. 18 is another example diagram of the connection relationship of points in a three-dimensional grid.

FIG19 is a flow chart of the decoding method provided in an embodiment of the present application.

FIG20 is a flow chart of the encoding method provided in an embodiment of the present application.

FIG. 21 is a diagram illustrating an example of the prediction and transformation process of the shift coefficients.

FIG. 22 is a diagram showing an example of the structure of the LOD layer.

FIG. 23 is a schematic diagram of the structure of a decoder provided in an embodiment of the present application.

FIG24 is a schematic diagram of the structure of a decoder provided in another embodiment of the present application.

FIG. 25 is a schematic diagram of the structure of an encoder provided in one embodiment of the present application.

FIG. 26 is a schematic diagram of the structure of an encoder provided in another embodiment of the present application.

DETAILED DESCRIPTION

In order to enable a more detailed understanding of the features and technical contents of the embodiments of the present application, the implementation of the embodiments of the present application is described in detail below in conjunction with the accompanying drawings. The attached drawings are for reference only and are not used to limit the embodiments of the present application.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

It should also be pointed out that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

Generally speaking, 3D animation content is represented based on keyframes, that is, each frame is a static mesh. Static meshes at different times have the same topological structure and different geometric structures. However, the amount of data of 3D dynamic meshes represented based on keyframes is extremely large, so how to effectively store, transmit and draw them has become a problem faced by the development of 3D dynamic meshes. In addition, the spatial scalability of the mesh needs to be supported for different user terminals (computers, notebooks, portable devices, mobile phones); different network bandwidths (broadband, narrowband, wireless) need to support the quality scalability of the mesh. Therefore, 3D dynamic mesh compression is a very critical issue.

A 3D mesh is a 3D object surface composed of several polygons in space. A polygon can be composed of vertices and edges. FIG. 1A shows a 3D mesh image, and FIG. 1B shows a partial magnified view of the 3D mesh image. As can be seen from FIG. 1A and FIG. 1B, a mesh surface is usually composed of multiple closed polygons.

The pixel distribution of a two-dimensional image is regular, so there is no need to record its geometric information (or position information) additionally. However, the distribution of the vertices in the mesh in three-dimensional space is random and irregular, and the way the polygons are formed requires additional recording. Therefore, for a three-dimensional mesh, it is necessary not only to record the position of the vertices in the mesh in space, but also to record the connection information of the polygons in the mesh in order to fully express a mesh image. As shown in Figure 2, the same number of vertices and vertex positions, due to different connection methods, form completely different surfaces.

In addition to the above information, since 3D grid images are usually encoded using existing 2D image/video encoding methods, To transform the 3D grid image from 3D space to 2D space, 3D grid encoding usually uses UV coordinates to define this transformation process.

Similar to two-dimensional images, each vertex may have corresponding attribute information. Attribute information is usually RGB color value to reflect the color of the object. For three-dimensional mesh images, in addition to color, the attribute information corresponding to each vertex is also commonly reflectance value, which reflects the surface material of the object. The attribute information of a three-dimensional mesh image can be stored in a two-dimensional image, and its mapping from two-dimensional to three-dimensional is specified by UV coordinates.

Therefore, the three-dimensional mesh data usually includes three-dimensional geometric position information (x, y, z), the connection relationship of the geometric position information triangles, texture coordinates (u, v) and the connection relationship of the texture coordinates, and an attribute map. Figure 3A is a three-dimensional mesh image, and the mesh data storage format shown in Figure 3B includes three-dimensional geometric position information, texture coordinates and connection information, and Figure 3C is the corresponding attribute map.

Current 3D dynamic mesh compression methods include space-time prediction methods, which improve compression efficiency by eliminating spatial and temporal correlations; principal component analysis (PCA)-based technology, which projects in the eigenvector space to concentrate energy; and wavelet-based methods, which support spatial scalability and quality scalability.

FIG4 is a diagram of the overall framework of mesh coding. FIG5A is a schematic diagram of the preprocessing process of a two-dimensional curve. FIG5B is a schematic diagram of the generation of a displacement coefficient. The preprocessing process of a three-dimensional mesh can be compared to the preprocessing process of a two-dimensional curve. At the encoding end, it is mainly divided into two parts: preprocessing and encoding. First, the base mesh and displacement coefficient can be generated by preprocessing. The preprocessing process includes: first, downsampling the original mesh to generate a simplified mesh (decimated mesh) with a greatly reduced number of vertices, or a base mesh. Then, the simplified mesh is subdivided, and the newly generated vertices are inserted on the edges of the simplified mesh to obtain a subdivided mesh or an initial mesh. Finally, for each vertex in the subdivided mesh, find the point in the original mesh that is closest to it, and calculate the displacement value of the two points. After preprocessing, the simplified mesh and the displacement coefficient are input into the encoder to generate a bitstream.

V-DMC encoding can be mainly divided into two categories: geometric position information encoding and attribute information encoding. Each frame file of the sequence basketball_player includes two files: basketball_player_fr0001_qp12_qt12.obj and basketball_player_fr0002.png. Among them, basketball_player_fr0001_qp12_qt12.obj contains four types of information: geometric position information (x, y, z), the connection relationship of geometric position information triangles, texture coordinates (u, v) and the connection relationship of texture coordinates. basketball_player_fr0002.png represents the attribute information of the current frame. In the current V-DMC encoder, the geometric position information is jointly encoded by DRACO and Video Codec (AVC, HEVC or VVC), and the texture information encoding is directly encoded by Video Codec. Therefore, the mesh geometric information encoding is introduced in detail below.

Geometric information can be divided into: encoding of position information (geometric position information and texture position information) and encoding of connection relationships (connection relationships of triangle patches of geometric position information and connection relationships of texture position information). The current V-DMC encoding is mainly divided into two encoding test conditions: intra-frame encoding and inter-frame encoding (low latency, currently there is no RA test environment).

Intra-frame geometry coding

1. Mesh preprocessing

a) As shown in Figures 5A and 5B, the two-dimensional curve preprocessing process is taken as an example. The connection relationship of the original grid contains a large number of points. Before encoding the grid geometry information, the grid geometry information is first quantized or simplified, and finally the corresponding simplified grid is obtained as the basic grid.

b) As shown in FIG6A and FIG6B , the quantization processing of the grid is performed based on the coordinates of the triangular patch. According to the connection relationship between the quantization points, the quantization processing is divided into the following two cases:

When two vertices belong to one edge before quantization, after quantization, all the triangles connected to the two vertices need to be connected together. As shown in FIG6A , this quantization processing method involves the disappearance of the previous triangles. If the two vertices do not belong to one edge, after quantization, only the boundaries of the two vertices need to be merged. As shown in FIG6B , this quantization processing method has no effect on the number of triangles.

c) In the whole process of mesh quantization based on triangle patch coordinates, the core issue is how to get the best vertex based on the previous vertex coordinates. The current V-DMC will select the best quantization point from the following four modes. Assuming that the vertex distribution before quantization is V1 and V2, and the vertex coordinates after quantization are V', there are the following quantization points: V1, V2, (V1+V2)/2 and Q ^{- 1} (V1+V2), where Q is the quantization matrix corresponding to the vertex coordinates of V1 and V2. Finally, the best quantization point is selected based on the distortion measure D before and after quantization.

2. Basic Grid Coding

a) After obtaining the base mesh, the DRACO encoder is used to encode the geometric information of the base mesh. The geometric information mainly includes: the connection relationship between the geometric position information and the geometric position information. The entire DRACO encoding process is as follows: first complete the encoding of the connection relationship, then encode the geometric position information of the point based on the connection relationship of the geometric position, and finally encode the texture position information based on the connection relationship and geometric position information.

b) Coding of connection relations. DRACO uses the "Edgebreaker Coding" scheme to encode the connection relations of the grid. As shown in Figure 7A:

Before encoding the connection relationship of the mesh, the vertices of the mesh are divided into five types, CLRSE, where each symbol means the following:

i.C: None of the triangles connected to the current vertex have been encoded;

ii.L: The left triangle connected to the current vertex completes the encoding;

iii.R: The right triangle connected to the current vertex completes the encoding;

iv.S: The left and right triangles connected to the current vertex have not been encoded;

v.E: The left and right triangles connected to the current vertex have been encoded.

Finally, the type of each vertex and the processing order of the vertices are encoded in a certain order, and the decoding end restores the geometric connection relationship of the mesh according to the processing order and type of the vertices.

c) Coding of geometric position information. After completing the coding of the vertex connection relationship, the geometric position information of each vertex is predictively coded based on the vertex connection relationship. The idea adopted in predictive coding is the "Parallelograms algorithm", as shown in Figure 7B:

A simple linear fit is performed using the three vertices adjacent to the current point to be encoded: the left vertex, the right vertex, and the opposite vertex:
pred _pos =(left+right)-opposite

d) After completing the point connection relationship and geometric position information, the texture coordinates are predicted and encoded based on the decoding and reconstruction of the two, as shown in Figure 7C. Similarly, assuming that the current vertex is C, the Left and Right vertices of the current point can be obtained according to the point connection relationship, and then the texture coordinates of the left and right vertices are used to predict the texture coordinates of the current vertex C.

3. Displacement encoding

a) First, after the encoding and reconstruction of the base mesh is completed, a certain partitioning algorithm will be used to partition the base mesh to obtain the initial reconstructed mesh. For details, please refer to the curve corresponding to the subdivided mesh in Figure 5A. The curve corresponding to the simplified mesh can be obtained by simple linear interpolation. The coordinates of the newly inserted point are obtained by linear interpolation based on the two vertices on the current boundary:

b) Secondly, the error Delta between the points of the subdivided mesh and the original mesh is calculated. The error Delta can be a point error in the world coordinate system. Finally, the Displacement value of each point is calculated using the error Delta between each point and the normal vector Norm of each point, as shown in Figure 8A.

The specific calculation method is as follows:
Displacement = Delta × Norm

c) After calculating the Displacement of each point, the lifting wavelet transform can be used to transform the spatial domain residual coefficients to the frequency domain to obtain the corresponding frequency domain residual coefficients.

d) Finally, the Packing algorithm is used to map the frequency domain residual coefficients of each point into a two-dimensional image in a certain order. The current V-DMC is arranged in the order of the Morton code, as shown in FIG8B .

e) Finally, the traditional Video Codec can be used to encode the two-dimensional image.

4. Recoloring

Recoloring is an algorithm on the encoder side. After the reconstruction of the geometric information on the encoder side is completed, the texture attribute information of the reconstructed mesh is recolored using the original geometric information, the original texture attribute information, and the reconstructed mesh geometric information.

Inter-frame geometry information coding

a) Similar to intra-frame geometric information coding, including geometric connection relationship and geometric position information coding. However, it should be noted that the inter-frame geometric position information coding only needs to encode the geometric position information (x, y, z) of the current Base mesh, and does not need to encode the connection relationship and texture position information (u, v). The specific reasons are as follows: If the current frame can be inter-frame coded, then the Base mesh of the reference frame of the current frame will be used at the encoding end to obtain the mesh information of the current frame. Therefore, the current frame and the reference frame have the same connection relationship and uv texture coordinates, but the geometric position information is different.

b) Based on a), we know that there is only an error in the geometric position information between the current frame and the reference frame, so the current V-DMC predicts and encodes the geometric position information of the current frame.

As shown in Figure 9, the black vertex in the middle is the vertex to be encoded/decoded. The current point is used in the reference frame to obtain the corresponding prediction point (similar to the same-position block in video encoding). Then, the neighborhood point of the current point (the MV of the vertices that have been encoded) is used to predict the MV of the current point. As shown below, assuming that the coordinates of the current point are pos and the coordinates of the corresponding same-position point are Pred_pos, the MV of the current point is calculated as:
MV＝Pos-Pred _pos

There are two prediction coding modes in the current V-DMC:

i. Directly encode the MV of the current point

ii. Use the neighborhood to predict the MV of the current point

At the encoding end, the rate-distortion optimization algorithm is used to obtain the optimal coding mode for each coding group (CG). The current V-DMC sets the maximum number of points for each CG to 16.

Coding of texture attribute information: The current V-DMC encodes texture attribute information directly using Video-Codec, such as AVC, HEVC, VVC or VV-Enc.

FIG10A is a schematic diagram of intra-frame coding. As shown in FIG10A , in the intra-frame encoder, a common static mesh encoder (Static Mesh Encoder) can be used to encode the simplified mesh to generate the corresponding bitstream (compressed base mesh bitstream). Next, the displacement coefficient is updated (update displacement) using the reconstructed simplified mesh. The updated displacement coefficient is subjected to wavelet transform (wavelet transform) and quantization (quantization) to obtain the displacement coefficient. After image packing (image packing), high efficiency video coding (HEVC) is used for encoding to generate a bitstream (compressed displacements bitstream) of the displacement coefficient. For attribute map encoding, the feature map is first transformed (texture transfer) according to the difference between the reconstructed geometric information and the original geometric information, and then padded (padding) and color space conversion (color space conversion) are performed, and then encoded using a video encoder (video coding) to form a compressed attribute bitstream.

Figure 10B is a schematic diagram of inter-frame coding. As shown in Figure 10B, the inter-frame encoder and intra-frame encoder processes are roughly the same, but the inter-frame encoder does not directly encode the simplified grid, but encodes the motion vector between the simplified grid of the current frame and the simplified grid of the reference frame (motion encoder), and generates the corresponding motion vector bitstream (compressed motion bitstream).

Common test conditions for MPEG DMC

1) There are two test conditions for MPEG DMC:

Condition 1: all intra geometry lossy, attributes lossy;

Condition 2: random access is lossy in geometry and attributes;

2) The general test sequences include five categories: Cat1-A, Cat1-B and Cat1-C, all of which contain geometric and color attribute information.

Next, the Displacement coding of V-DMC is introduced in more detail.

Encoding algorithm:

At the encoding end, the basic grid is first iteratively divided according to a certain algorithm to obtain the geometric position information of the points in the subdivided grid. The specific division algorithm is consistent with the background introduction, that is, linear interpolation is performed using the vertices on each boundary to obtain the geometric position information of the points in the subdivided grid. Assuming that the entire division is iterated N times, the N iterative division process can be LOD divided to obtain multiple levels, as shown in Figure 11.

After obtaining the geometric position information of the midpoint of the subdivided grid, the geometric position information of the midpoint of the subdivided grid and the geometric position information of the midpoint of the original grid can be used to perform error calculation to obtain the shift coefficient of the midpoint of the subdivided grid. The LOD hierarchy of the shift coefficient is shown in FIG12.

Secondly, a lifting transform is performed based on the LOD spatial structure. The lifting transform includes two steps: prediction and update. The prediction algorithm is as follows:

The update algorithm is as follows:

Finally, the transformed coefficients are quantized, and the quantized coefficients are reorganized, as shown in FIG13 .

As shown in FIG14 , V-DMC can reorganize coefficients based on blocks, and the size of each block can be, for example, 16×16. The coefficients within each block can be arranged according to the Morton code to obtain the corresponding two-dimensional image. After completing the above operations, the two-dimensional image can be encoded using Video-Codec.

Decoding algorithm:

First, use Video-Codec to decode and reconstruct the two-dimensional image to obtain the corresponding two-dimensional image. Secondly, restore the transform coefficient corresponding to each point by coefficient reorganization. Finally, use the inverse transform of the lifting transform to restore the shift coefficient of each point. After obtaining the shift coefficient of each point, as shown in Figure 15, the geometric position information of the basic grid and the shift coefficient can be used to reconstruct and restore the geometric position information corresponding to the current grid.

The previous article describes in detail the encoding and decoding process of the geometric information of points in a three-dimensional grid. How to improve the encoding and decoding efficiency of geometric information is a technical problem that this application needs to solve. After research, it was found that the importance of different points in a three-dimensional grid may be different. The difference in importance between different points can be reflected in the different connection relationships between different points, or the different number of neighborhood points of different points. When encoding and decoding the shift coefficients of points in a three-dimensional grid, the related technology did not fully consider the difference in importance of different points, thereby limiting the efficiency of several The encoding and decoding efficiency of any information.

For example, in the V-DMC provided by the related art, the basic grid is first divided to determine the geometric information of the subdivided grid; then, the difference between the original geometric information of the three-dimensional grid and the geometric information of the subdivided grid is calculated to determine the shift coefficients of the points in the three-dimensional grid; then, the lifting transform is used to transform the shift coefficients from the spatial domain to the frequency domain, thereby completing the encoding of the geometric information of the V-DMC. The lifting transform scheme provided by the related art includes two processes: prediction and transformation (or update). Among them, the prediction scheme uses two neighboring points that are collinear with the current point for mean prediction, and the transformation scheme uses the residual coefficients of the neighboring point set of the current point (the points predicted based on the current point) to update the shift coefficient of the current point. Based on such a coding scheme, the spatial redundancy between adjacent points is removed by the prediction scheme, and the frequency domain redundancy between adjacent points is further removed by the transformation scheme, which can improve the coding efficiency of the geometric information of the three-dimensional grid to a certain extent. However, in the prediction scheme provided by the related art, the prediction weight is fixed at 0.5, and the update weight is usually fixed at 0.125 (indicated by the syntax element in the adaptation parameter set (aps)). The use of fixed weights for prediction and/or transformation does not fully consider the differences in importance between different points, thereby limiting the coding efficiency of the geometric information of V-DMC.

The connection relationship between different points in the grid and the surrounding points may be different, so the number of neighborhood points of the same point may also be different. If the number of neighborhood points of two points is different, the number of points predicted based on the two points may be different, so the importance of the two points in the grid may be different. Taking Figure 16 as an example, the number of neighborhood points collinear with point 1 is 6, indicating that the shift coefficients of the 6 points are predicted based on the shift coefficient of point 1, so the residual coefficients of the 6 points can be used to update the shift coefficient of point 1. Similarly, the number of neighborhood points collinear with point 2 is 5, indicating that the shift coefficients of the 5 points are predicted based on the shift coefficient of point 2, so the residual coefficients of the 5 points can be used to update the shift coefficient of point 1. By comparing point 1 and point 2, it can be seen that the importance of the two in the grid is different. If a fixed prediction weight and/or update weight is used, it is not reasonable. Therefore, the embodiment of the present application introduces the weight corresponding to the point, and predicts and/or updates based on the weight corresponding to the point (the weight corresponding to the point can be introduced in the prediction process, the weight corresponding to the point can also be introduced in the update process, or the weight corresponding to the point can be introduced in both the prediction and update processes). Predicting and/or updating based on the weight corresponding to the point helps to improve the coding efficiency of the geometric information of V-DMC.

The above describes the problems existing in the prediction and/or transformation process of the shift coefficients and possible improvements. The following describes the problems existing in the quantization process of the shift coefficients and possible improvements.

The V-DMC coding scheme provided by the related art needs to quantize the transformation coefficients after transforming the shift coefficients of each point. V-DMC performs layer-by-layer quantization based on the spatial structure of LOD. As shown in Figure 17, the quantization parameter QP _Lvl of each layer can be calculated as follows:
QP _Lvl ＝liftingQP×Scale ^Lvl-1

Among them, Lvl represents the index of each layer of LOD, liftingQP is the initial quantization parameter, and Scale is the adjustment parameter of each layer of LOD based on liftingQP.

After obtaining the quantization parameters of each layer, the shift coefficients of each point can be adaptively quantized in the following way:

Among them, bitDepth represents the effective bit depth of the shift coefficient in the current geometric coding, Res represents the transform coefficient of the current point to be encoded, and QP _Lvl represents the quantization parameter of the current LOD layer.

At the decoding end, the quantization coefficient of each point is recovered by entropy decoding and parsing, and the quantization parameter QP _Lvl of each LOD layer is obtained in the same way as the encoding end. The shift coefficient of the current point is dequantized based on the quantization parameter. The specific calculation method is as follows:

In summary, the quantization parameter QP _Lvl of different LOD layers is obtained by using the initial quantization parameter liftingQP and the quantization adjustment parameter Scale of different LOD layers, which can improve the coding efficiency of the geometric information of the three-dimensional grid to a certain extent. This is because the shift coefficient is adaptively transformed and encoded using the lifting transformation based on the LOD spatial structure, resulting in the shift coefficient of the LOD high layer being a low-frequency coefficient, while the shift coefficient of the LOD low layer is a high-frequency coefficient. Therefore, based on the above quantization scheme, adaptive quantization coding can be implemented for different LOD layers. However, for points on the same layer, since the spatial connection relationship of each point is inconsistent, the importance of each point in the lifting transformation process will be inconsistent. However, the above quantization coding scheme does not fully take this into account, thereby limiting the coding and decoding efficiency of the geometric information of the three-dimensional grid to a certain extent.

Referring to Figure 18, the three points P1, P2 and P3 are at the same level LOD, the neighborhood point set of P1 (i.e., the points predicted based on P1) includes 6 neighborhood points, the neighborhood point set of P2 (i.e., the points predicted based on P2) includes 6 neighborhood points, and the neighborhood point set of P3 (i.e., the points predicted based on P3) includes only 5 neighborhood points. Then for these three points, the importance is P1=P2>P3. The higher the importance, the greater the influence of the shift coefficient of the point on the reconstruction quality of the entire grid. Therefore, when quantizing the shift coefficients with high importance, the quantization degree of the shift coefficient can be reduced as much as possible (i.e., reducing the quantization parameter QP); and when quantizing the shift coefficients with low importance, the quantization degree can be appropriately increased (i.e., increasing the quantization parameter QP). In this way, the final reconstruction quality can be guaranteed, and the encoding bit rate of geometric information can be improved. Based on such a quantization coding scheme, the importance of different points is fully considered, so that the quality of the reconstruction can be improved. Coding and decoding efficiency of geometric information of V-DMC.

The following text, in conjunction with Figure 19, first describes in detail the implementation method of the embodiment of the present application in the decoding process from the perspective of the decoding end.

Figure 19 is a flowchart of a decoding method provided in an embodiment of the present application. The method of Figure 19 may be executed by a decoder. The decoder may be a decoder supporting V-DMC.

19, in step S1910, the code stream is parsed to determine the quantization coefficients of the points in the three-dimensional grid. It should be understood that the quantization coefficients of the points mentioned in the embodiments of the present application refer to the quantization coefficients corresponding to the shift coefficients of the points.

In step S1920, the quantization coefficients of the points in the three-dimensional grid are dequantized to determine the dequantization coefficients of the points in the three-dimensional grid.

In step S1930, the inverse quantization coefficients of the points in the first layer are inversely transformed to determine the shift coefficients of the points in the first layer. The points in the three-dimensional grid belong to multiple LOD layers, and the first layer mentioned in step S1930 can be any LOD layer among the multiple LOD layers.

In step S1940, the shift coefficients of the points in the second layer are predicted according to the shift coefficients of the points in the first layer. The second layer is an LOD layer adjacent to the first layer among the multiple LOD layers, and the second layer is a layer below the first layer. Steps S1930 and S1940 may belong to two sub-processes of the inverse transformation of the lifting transformation.

FIG. 19 illustrates the decoding process of the first layer and the second layer as an example. Each LOD layer in the multiple LOD layers can be processed in a similar manner to obtain the shift coefficients of the points in the three-dimensional grid. Then, the reconstructed geometric information of the points in the three-dimensional grid can be determined based on the shift coefficients of the points in the three-dimensional grid and the geometric information of the points in the subdivided grid (or the initial reconstructed grid) (the geometric information of the subdivided grid can also be called the initial geometric information of the three-dimensional grid). For example, according to the shift coefficients of the points in the three-dimensional grid, the geometric information of the corresponding points in the subdivided grid can be shifted to determine the reconstructed geometric information of the points in the three-dimensional grid. The geometric information of the points in the subdivided grid can be obtained based on the division of the basic grid. For example, the code stream can be parsed first to determine the geometric information of the points in the basic grid; then, according to the geometric information of the points in the basic grid, the basic grid can be divided to determine the geometric information of the points in the subdivided grid.

According to Figure 19, the complete decoding operation includes inverse quantization operation, inverse transformation operation, prediction operation, etc. In order to improve the decoding efficiency of the geometric information of the three-dimensional grid, one or more of the inverse quantization operation, inverse transformation operation, and prediction operation can be optimized based on the weights corresponding to the points in the three-dimensional grid (or the spatial connection relationship of the points). The following examples illustrate the implementation methods provided in the embodiments of the present application from the three perspectives of inverse quantization operation, inverse transformation operation, and prediction operation. It should be understood that the embodiments of the present application can optimize only one of inverse quantization, inverse transformation, and prediction; or, the embodiments of the present application can optimize any two of inverse quantization, inverse transformation, and prediction, thereby further improving the decoding efficiency of the geometric information of the three-dimensional grid; or, the embodiments of the present application can optimize inverse quantization, inverse transformation, and prediction at the same time, thereby further improving the decoding efficiency of the geometric information of the three-dimensional grid.

For ease of understanding, the inverse transformation operation of the points in the three-dimensional grid is described below from the perspective of the first point to be decoded. The first point to be decoded may be any point to be decoded in the first layer. The neighborhood point of the first point to be decoded is called the first neighborhood point. The first neighborhood point may be a point in the next layer of the LOD layer where the first point to be decoded is located. As an example, the first neighborhood point is a point predicted based on the first point to be decoded.

Step S1930 in Figure 19, i.e., inversely transforming the inverse quantization coefficients of the points in the first layer, may include: inversely transforming the inverse quantization coefficients of the first point to be decoded in the first layer according to the inverse quantization coefficients of the first neighborhood points and the updated weights. In the related art, the updated weights use a fixed value (i.e., 0.125). Different from the related art, in the embodiment of the present application, the updated weights are determined based on the weights corresponding to the first neighborhood points. The updated weights are determined based on the weights corresponding to the neighborhood points, so that the points with higher importance have a greater impact on the inverse transformation results of the points to be decoded, which makes the updated results more accurate and helps to improve the decoding efficiency of the geometric information of the points in the three-dimensional grid.

The weight corresponding to the first neighborhood point can be determined based on the spatial connection relationship of the first neighborhood point in the three-dimensional grid. For example, the weight corresponding to the first neighborhood point can be related to the number of the neighborhood points (hereinafter referred to as the second neighborhood points) of the first neighborhood point. Determining the weight corresponding to the point based on the number of neighborhood points makes full use of the connection relationship information of the point in the three-dimensional grid, so that the weight of the determined point is more reasonable.

The embodiment of the present application does not specifically limit the definition of the second neighborhood point. For example, the second neighborhood point can be a point in the same LOD layer as the first neighborhood point. Alternatively, the second neighborhood point can be a point in the next layer of the LOD layer where the first neighborhood point is located. Alternatively, the second neighborhood point can include points in the above two LOD layers at the same time. As an example, the second neighborhood point is a point predicted based on the first neighborhood point.

In some implementations, the weight corresponding to the first neighborhood point may be equal to the number of the second neighborhood points. For example, the first neighborhood points include point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points, then the weight of point 1 may be 5, and the weight of point 2 may be 6.

In some implementations, the weight corresponding to the first neighborhood point can be determined based on the product of the number of second neighborhood points and the first prediction weight. For example, the weight corresponding to the first neighborhood point can be equal to the product of the number of second neighborhood points and the first prediction weight. The first prediction weight mentioned here can be a predefined fixed value. For example, the value of the first prediction weight can be 0.5. When determining the weight corresponding to the point, comprehensively considering the prediction weight can make the determined weight more reasonable.

Exemplarily, the weight corresponding to the first neighborhood point may be determined based on the following formula:

In the above formula, weight represents the weight corresponding to any point in the first neighborhood point, and predWeight is the first prediction weight (the value can be 0.5). P represents the second neighborhood point, and weight[i] represents the initial weight corresponding to the i-th point in the second neighborhood point. The initial weight corresponding to the i-th point can be set to 1. For example, if the first neighborhood point includes point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points, then based on the above formula, the weight of point 1 can be 2.5, and the weight of point 2 can be 3 (the value of the first prediction weight is 0.5).

In some implementations, the weight corresponding to the first neighborhood point can be determined based on the weight corresponding to the LOD layer where the first neighborhood point is located. For example, the weight corresponding to the first neighborhood point can be equal to the weight corresponding to the LOD layer where the first neighborhood point is located. The weight corresponding to the LOD layer can be understood as a level weight. The level weight of a LOD layer can be determined, for example, based on the ratio of the number of points in the LOD layer to the total number of points to be decoded.

As mentioned above, the update weight can be determined based on the weight corresponding to the first neighborhood point. As a possible implementation, the update weight can be equal to the weight corresponding to the first neighborhood point. In another possible implementation, the update weight can be determined based on the weight corresponding to the first neighborhood point and the first prediction weight (the first prediction weight can be a predefined fixed value, for example, the value of the first prediction weight can be 0.5). For example, the update weight can be determined based on the product of the weight corresponding to the first neighborhood point and the first prediction weight, or the update weight can be equal to the product of the weight corresponding to the first neighborhood point and the first prediction weight. When determining the update weight, considering the prediction weight can make the weight distribution of the prediction and update process more reasonable.

Exemplarily, the update weight may be determined based on the following formula:
updateWeight _i = predWeight × weight _i

In the above formula, updateWeight _i represents the update weight corresponding to the i-th point in the first neighborhood point. predWeight represents the first prediction weight, and the value of the first prediction weight can be 0.5. weight _i represents the weight corresponding to the first neighborhood point. The weight corresponding to the first neighborhood point can be determined using the implementation method described above.

The above describes in detail the weight corresponding to the first neighborhood point and the method for determining the updated weight. Based on the determination of the weight corresponding to the first neighborhood point and the updated weight, the inverse quantization coefficient of the first point to be decoded can be inversely transformed according to the following formula:

In the above formula, Signal represents the shift coefficient of the first point to be decoded. UpdateWeight _i represents the update weight corresponding to the i-th point in the first neighborhood point. The calculation method of updateWeight _i can be found in the previous text. P represents the point set formed by the first neighborhood point, and i represents the i-th point in the first neighborhood point. Signal _i represents the inverse quantization coefficient of the first neighborhood point.

In addition to the solution of performing inverse transformation based on the weight corresponding to the point, the embodiment of the present application can also apply other inverse transformation solutions. For example, the inverse transformation can be performed based on the adjustment parameter (Scale) of the updated weight, as shown in the following formula:

In the above formula, updateWeight represents the update weight (can be a fixed value, such as 0.125). ^{Scale ni-1} is the adjustment parameter of the update weight, n represents the number of LOD layers, and i represents the index of the LOD layer. P represents the point set formed by the first neighborhood point, and i represents the i-th point in the first neighborhood point. _{signal i} represents the inverse quantization coefficient of the first neighborhood point.

The above text describes in detail the inverse transformation operation of the points in the three-dimensional grid . The following text describes the prediction operation of the points in the three-dimensional grid from the perspective of the second point to be decoded. The second point to be decoded can be any point to be decoded in the second layer. The neighborhood point of the second point to be decoded is called the third neighborhood point. The embodiment of the present application does not specifically limit the definition method of the third neighborhood point. For example, the third neighborhood point can be a point in the same LOD layer as the second point to be decoded. Alternatively, the third neighborhood point can be a point in the next layer of the LOD layer where the second point to be decoded is located. Alternatively, the third neighborhood point can include points in the above two LOD layers at the same time. As an example, the third neighborhood point can be a point in the first layer (the upper layer of the LOD layer where the second point to be decoded is located) that is collinear with the second point to be decoded.

Step S1940 in Figure 19, i.e. predicting the shift coefficients of points in the second layer based on the shift coefficients of points in the first layer, may include: predicting the second point to be decoded based on the shift coefficients of the third neighborhood points and the second prediction weight. In the related art, the prediction weight adopts a fixed value (i.e. 0.5). Different from the related art, in the embodiment of the present application, the second prediction weight is determined based on the weight corresponding to the third neighborhood point. The larger the weight corresponding to a point in the third neighborhood point, the greater the influence of the point on the prediction result of the second point to be decoded. The prediction weight is determined based on the weight corresponding to the neighborhood point, which can make the points with higher importance have a greater influence on the prediction result of the point to be decoded, which will make the prediction process more reasonable and help improve the decoding efficiency of the geometric information of the points in the three-dimensional grid.

The weight corresponding to the third neighborhood point can be determined based on the spatial connection relationship of the third neighborhood point in the three-dimensional grid. For example, the weight corresponding to the third neighborhood point can be related to the number of neighboring points of the third neighborhood point (hereinafter referred to as the fourth neighborhood point). Determining the weight corresponding to the point based on the number of neighborhood points makes full use of the connection relationship information of the point in the three-dimensional grid, which can make the weight of the determined point more reasonable.

The embodiment of the present application does not specifically limit the definition of the fourth neighboring point. For example, the fourth neighboring point can be located at The fourth neighborhood point may be a point in the same LOD layer. Alternatively, the fourth neighborhood point may be a point in the next layer of the LOD layer where the third neighborhood point is located. Alternatively, the fourth neighborhood point may include points in both LOD layers. As an example, the fourth neighborhood point is a point predicted based on the third neighborhood point.

In some implementations, the weight corresponding to the third neighborhood point may be equal to the number of the fourth neighborhood points. For example, the third neighborhood points include point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points, then the weight of point 1 may be 5, and the weight of point 2 may be 6.

In some implementations, the weight corresponding to the third neighborhood point can be determined based on the product of the number of fourth neighborhood points and the first prediction weight. For example, the weight corresponding to the third neighborhood point can be equal to the product of the number of fourth neighborhood points and the first prediction weight. The first prediction weight mentioned here can be a predefined fixed value. For example, the value of the first prediction weight can be 0.5. When determining the weight corresponding to the point, considering the prediction weight can make the determined weight more reasonable.

Exemplarily, the weight corresponding to the third neighborhood point may be determined based on the following formula:

In the above formula, weight represents the weight corresponding to any point in the third neighborhood point, and predWeight is the first prediction weight (the value can be 0.5). P represents the fourth neighborhood point, and weight[i] represents the initial weight corresponding to the i-th point in the fourth neighborhood point. The initial weight corresponding to the i-th point can be set to 1. For example, if the third neighborhood point includes point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points, then based on the above formula, the weight of point 1 can be 2.5, and the weight of point 2 can be 3 (the first prediction weight is 0.5).

In some implementations, the weight corresponding to the third neighborhood point can be determined based on the weight corresponding to the LOD layer where the third neighborhood point is located. For example, the weight corresponding to the third neighborhood point can be equal to the weight corresponding to the LOD layer where the third neighborhood point is located. The weight corresponding to the LOD layer can be understood as a level weight. The level weight of a LOD layer can be determined, for example, based on the ratio of the number of points in the LOD layer to the total number of points to be decoded.

As mentioned above, the second prediction weight can be determined based on the weight corresponding to the third neighborhood point. For example, the second prediction weight can be equal to the weight corresponding to the third neighborhood point. Alternatively, the second prediction weight can be based on the product of the weight corresponding to the third neighborhood point and a certain parameter.

As mentioned above, the second point to be decoded can be predicted according to the shift coefficient of the third neighborhood point and the second prediction weight. A possible prediction method for the second point to be decoded is given below.

Exemplarily, the second to-be-decoded point may be predicted based on the following formula:

In the above formula, Signal represents the prediction result of the second point to be decoded, and Neigh ₁ and Neigh ₂ represent the shift coefficients of the two neighboring points collinear with the second point to be decoded. _{Neigh 1} and Neigh ₂ are located in the upper layer of the LOD layer where the second point to be decoded is located. predWeight ₁ and predWeight ₂ represent the prediction weights corresponding to Neigh ₁ and Neigh ₂ (i.e., the second prediction weights mentioned above).

predWeight ₁ and predWeight ₂ can be calculated using the following formula:
predWeight ₁ = weight ₁
predWeight ₂ = weight ₂

sumWeight can be calculated based on the following formula:
sumWeight＝predWeight ₁ +predWeight ₂

The above describes the inverse transformation operation and prediction operation of the points in the three-dimensional grid in detail, and the following describes the inverse quantization operation of the points in the three-dimensional grid from the perspective of the third point to be decoded. The third point to be decoded can be any point to be decoded in the three-dimensional grid.

Step S1920 in Figure 19, i.e., dequantizing the quantization coefficients of the points in the three-dimensional grid, may include: dequantizing the quantization coefficients of the third point to be decoded according to the weight corresponding to the third point to be decoded. When dequantizing the points to be decoded in the three-dimensional grid, the embodiment of the present application considers the weight corresponding to the point to be decoded. For example, a suitable quantization parameter may be selected according to the weight corresponding to the point to be decoded; or, the dequantization result may be processed differently according to the weight corresponding to the point to be decoded. Considering the weight of the point to be decoded in the dequantization process can make the dequantization result more reasonable, which helps to improve the decoding efficiency of the geometric information of the points in the three-dimensional grid.

The weight corresponding to the third point to be decoded can be determined based on the spatial connection relationship of the third point to be decoded in the three-dimensional grid. For example, the weight corresponding to the third point to be decoded can be related to the number of neighboring points (hereinafter referred to as the fifth neighboring points) of the third point to be decoded. Determining the weight corresponding to the point based on the number of neighboring points makes full use of the connection relationship information of the point in the three-dimensional grid, which can make the weight of the determined point more reasonable.

The embodiment of the present application does not specifically limit the definition of the fifth neighborhood point. For example, the fifth neighborhood point may be a point in the same LOD layer as the third point to be decoded. Alternatively, the fifth neighborhood point may be a point in the next layer of the LOD layer where the third point to be decoded is located. Alternatively, the fifth neighborhood point may include points in both LOD layers. As an example, the fifth neighborhood point is a point predicted based on the third point to be decoded.

In some implementations, the weight corresponding to the third point to be decoded may be equal to the number of the fifth neighboring points. For example, the third point to be decoded includes point 1 and point 2, point 1 includes 5 neighboring points, and point 2 includes 6 neighboring points. Then the weight of point 1 may be 5, and the weight of point 2 may be It can be 6.

In some implementations, the weight corresponding to the third point to be decoded may be determined based on the product of the number of the fifth neighborhood points and the first prediction weight. For example, the weight corresponding to the third point to be decoded may be equal to the product of the number of the fifth neighborhood points and the first prediction weight. The first prediction weight mentioned here may be a predefined fixed value. For example, the value of the first prediction weight may be 0.5. When determining the weight corresponding to the point, considering the prediction weight may make the determined weight more reasonable.

Exemplarily, the weight corresponding to the third to-be-decoded point may be determined based on the following formula:

In the above formula, weight represents the weight corresponding to any point in the third point to be decoded, and predWeight is the first predicted weight (the value can be 0.5). P represents the fifth neighborhood point, and weight[i] represents the initial weight corresponding to the i-th point in the fifth neighborhood point. The initial weight corresponding to the i-th point can be set to 1. For example, the third point to be decoded includes point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points. Based on the above formula, the weight of point 1 can be 2.5, and the weight of point 2 can be 3 (the first predicted weight is 0.5).

As mentioned above, the quantization coefficient of the third point to be decoded can be dequantized according to the weight corresponding to the third point to be decoded. A possible implementation manner is given below.

For example, in some implementations, the quantization coefficient of the third point to be decoded can be inversely quantized according to the first quantization parameter. The first quantization parameter can be determined based on the weight corresponding to the third point to be decoded. That is, the quantization parameter can be adaptively set for the point to be decoded according to the weight corresponding to the point to be decoded. For example, if the weight corresponding to the third point to be decoded is large, it means that the importance of the third point to be decoded is high. For points to be decoded with a high degree of importance, the value of the first quantization parameter can be small (indicating that the degree of quantization of the point to be decoded is low). For another example, if the weight corresponding to the third point to be decoded is small, it means that the degree of quantization of the third point to be decoded is low. For points to be decoded with a low degree of importance, the value of the first quantization parameter can be large (indicating that the degree of quantization of the point to be decoded is high). According to the design method of the above quantization parameters, the reconstruction quality can be guaranteed, thereby improving the decoding efficiency of the geometric information of the three-dimensional grid.

The first quantization parameter can be directly determined based on the weight corresponding to the third point to be decoded. For example, a mapping relationship between the weight corresponding to the point to be decoded and the quantization parameter can be pre-established, and the first quantization parameter can be determined based on the mapping relationship. Alternatively, the first quantization parameter can also be determined based on the weight corresponding to the third point to be decoded and the second quantization parameter. The second quantization parameter mentioned here can be the quantization parameter corresponding to the target LOD layer (i.e., the LOD layer to which the third point to be decoded belongs). The second quantization parameter can be determined based on the initial quantization parameter and the quantization adjustment parameter.

Exemplarily, the second quantization parameter may be determined based on the following formula:
QP _Lvl ＝liftingQP×Scale ^Lvl

In the above formula, Lvl represents the index of the target LOD layer, QP _Lvl is the second quantization parameter, liftingQP is the initial quantization parameter, and Scale is the quantization adjustment parameter.

A more specific example of inverse quantization of the third point to be decoded is given below.

For example, the third to-be-decoded point may be dequantized based on the following formula:

In the above formula, _InvResi represents the inverse quantization result of the third point to be decoded, _Resi represents the quantization coefficient of the third point to be decoded, bitDepth represents the effective bit depth of the shift coefficient, QP _Lvl represents the quantization parameter corresponding to the target LOD layer (the LOD layer where the third point to be decoded is located), and weight _i represents the weight corresponding to the third point to be decoded (the method for determining the weight corresponding to the third point to be decoded can be found in the previous text and will not be described in detail here).

The decoding method provided by the embodiment of the present application is described in detail above in conjunction with Figure 19. The encoding method provided by the embodiment of the present application is described in detail below in conjunction with Figure 20.

Figure 20 is a flow chart of a coding method provided in an embodiment of the present application. The method of Figure 20 may be performed by an encoder. The encoder may be an encoder supporting V-DMC.

Referring to Fig. 20, in step S2010, LOD division is performed on points in the three-dimensional grid to determine multiple LOD layers. The multiple LOD layers include an adjacent first layer and a second layer, where the first layer is a layer above the second layer.

In step S2020, the shift coefficients of the points in the second layer are predicted to determine the residual coefficients of the points in the second layer. It should be understood that the residual coefficients of the points mentioned in the embodiments of the present application refer to the residual coefficients corresponding to the shift coefficients of the points, or the predicted residuals of the shift coefficients.

In step S2030, the shift coefficients of the points in the first layer are transformed according to the residual coefficients of the points in the second layer. Steps S2020 and S2030 may belong to two sub-processes of the lifting transformation.

It should be understood that the shift coefficient of the point in the three-dimensional mesh can be determined based on the difference between the geometric information of the point in the subdivided mesh and the original geometric information of the three-dimensional mesh. For example, the base mesh can be divided according to the geometric information of the point in the base mesh to determine the geometric information of the point in the subdivided mesh; then, the shift coefficient of the point in the three-dimensional mesh can be determined according to the geometric information of the point in the subdivided mesh and the original geometric information of the three-dimensional mesh.

FIG. 20 is an example of the encoding process of the first layer and the second layer. Each LOD layer in the multiple LOD layers can be processed in a similar manner to obtain the transformation coefficients of the points in the three-dimensional grid. In addition, in some implementations, FIG. 20 may further include step S2040, i.e., quantizing the transformation coefficients of the points in the three-dimensional grid. The quantization of the transformation coefficients of the points in the three-dimensional grid can also be performed layer by layer.

According to Figure 20, the complete coding operation includes prediction operation, transformation operation and quantization operation. In order to improve the coding efficiency of the geometric information of the three-dimensional grid, one or more of the prediction operation, transformation operation and quantization operation can be optimized based on the weights of the points in the three-dimensional grid (or the spatial connection relationship of the points). The following examples illustrate the optimization scheme provided in the embodiment of the present application from the three perspectives of prediction operation, transformation operation and quantization operation. It should be understood that the embodiment of the present application can optimize only one of prediction, transformation and quantization; or, the embodiment of the present application can optimize any two of prediction, transformation and quantization, thereby further improving the coding efficiency of the geometric information of the three-dimensional grid; or, the embodiment of the present application can optimize prediction, transformation and quantization at the same time, thereby further improving the coding efficiency of the geometric information of the three-dimensional grid.

The following describes the transformation operation of points in the three-dimensional grid from the perspective of the first point to be encoded. The first point to be encoded may be any point to be encoded in the first layer. The neighborhood point of the first point to be encoded is called the first neighborhood point. The first neighborhood point may be a point in the next layer of the LOD layer where the first point to be encoded is located. As an example, the first neighborhood point is a point predicted based on the first point to be encoded.

Step S2030 in Figure 20, i.e., transforming the shift coefficients of the points in the first layer according to the residual coefficients of the points in the second layer, may include: transforming the shift coefficients of the first point to be encoded according to the residual coefficients of the first neighborhood points and the updated weights. In the related art, the updated weights use a fixed value (i.e., 0.125). Different from the related art, in the embodiment of the present application, the updated weights are determined based on the weights corresponding to the first neighborhood points. The updated weights are determined based on the weights corresponding to the neighborhood points, which can make the transformation results more reasonable and help improve the coding efficiency of the geometric information of the points in the three-dimensional grid.

The weight corresponding to the first neighborhood point can be determined based on the spatial connection relationship of the first neighborhood point in the three-dimensional grid. For example, the weight corresponding to the first neighborhood point can be related to the number of the neighborhood points (hereinafter referred to as the second neighborhood points) of the first neighborhood point. Determining the weight corresponding to the point based on the number of neighborhood points makes full use of the connection relationship information of the point in the three-dimensional grid, which can make the weight of the determined point more reasonable.

The embodiment of the present application does not specifically limit the definition of the second neighborhood point. For example, the second neighborhood point can be a point in the same LOD layer as the first neighborhood point. Alternatively, the second neighborhood point can be a point in the next layer of the LOD layer where the first neighborhood point is located. Alternatively, the second neighborhood point can include points in the above two LOD layers at the same time. As an example, the second neighborhood point is a point predicted based on the first neighborhood point.

In some implementations, the weight corresponding to the first neighborhood point may be equal to the number of the second neighborhood points. For example, the first neighborhood points include point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points, then the weight of point 1 may be 5, and the weight of point 2 may be 6.

In some implementations, the weight corresponding to the first neighborhood point can be determined based on the product of the number of second neighborhood points and the first prediction weight. For example, the weight corresponding to the first neighborhood point can be equal to the product of the number of second neighborhood points and the first prediction weight. The first prediction weight mentioned here can be a predefined fixed value. For example, the value of the first prediction weight can be 0.5. When determining the weight corresponding to the point, considering the prediction weight can make the determined weight more reasonable.

Exemplarily, the weight corresponding to the first neighborhood point may be determined based on the following formula:

In the above formula, weight represents the weight corresponding to any point in the first neighborhood point, and predWeight is the first prediction weight (the value can be 0.5). P represents the second neighborhood point, and weight[i] represents the initial weight corresponding to the i-th point in the second neighborhood point. The initial weight corresponding to the i-th point can be set to 1. For example, if the first neighborhood point includes point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points, then based on the above formula, the weight of point 1 can be 2.5, and the weight of point 2 can be 3 (the first prediction weight is 0.5).

In some implementations, the weight corresponding to the first neighborhood point can be determined based on the weight corresponding to the LOD layer where the first neighborhood point is located. For example, the weight corresponding to the first neighborhood point can be equal to the weight corresponding to the LOD layer where the first neighborhood point is located. The weight corresponding to the LOD layer can be understood as a level weight. The level weight of a LOD layer can be determined, for example, based on the ratio of the number of points in the LOD layer to the total number of points to be decoded.

As mentioned above, the update weight can be determined based on the weight corresponding to the first neighborhood point. As a possible implementation, the update weight can be equal to the weight corresponding to the first neighborhood point. In another possible implementation, the update weight can be determined based on the weight corresponding to the first neighborhood point and the first prediction weight (the first prediction weight can be a predefined fixed value, for example, the value of the first prediction weight can be 0.5). For example, the update weight can be determined based on the product of the weight corresponding to the first neighborhood point and the first prediction weight, or the update weight can be equal to the product of the weight corresponding to the first neighborhood point and the first prediction weight. When determining the update weight, considering the prediction weight can make the weight distribution of the prediction and update process more reasonable.

Exemplarily, the update weight may be determined based on the following formula:
updateWeight _i = predWeight × weight _i

In the above formula, updateWeight _i represents the update weight corresponding to the i-th point in the first neighborhood point. predWeight represents the first prediction weight, and the value of the first prediction weight can be 0.5. weight _i represents the weight corresponding to the first neighborhood point. The weight corresponding to the first neighborhood point can be determined by any of the above implementation methods.

The above describes in detail the weight corresponding to the first neighborhood point and the method for determining the updated weight. Based on the determination of the weight corresponding to the first neighborhood point and the updated weight, the shift coefficient of the first point to be encoded can be transformed according to the following formula:

In the above formula, Signal represents the shift coefficient of the first point to be encoded. UpdateWeight _i represents the update weight corresponding to the i-th point in the first neighborhood point. The calculation method of updateWeight _i can be found in the previous text. P represents the point set formed by the first neighborhood point, and i represents the i-th point in the first neighborhood point. Signal _i represents the residual coefficient (shift residual coefficient) of the first neighborhood point.

In addition to the scheme of transforming based on the weight corresponding to the point, the embodiment of the present application does not exclude other inverse transformation schemes. For example, the transformation can be performed based on the adjustment parameter (Scale) of the updated weight, as shown in the following formula:

In the above formula, updateWeight represents the update weight (can be a fixed value, such as 0.125). ^{Scale ni-1} is the adjustment parameter of the update weight, n represents the number of LOD layers, and i represents the index of the LOD layer. P represents the point set formed by the first neighborhood point, and i represents the i-th point in the first neighborhood point. signal _i represents the residual coefficient (shift residual coefficient) of the first neighborhood point.

The above text describes in detail the transformation operation of the points in the three-dimensional grid. The following text describes the prediction operation of the points in the three-dimensional grid from the perspective of the second point to be encoded. The second point to be encoded can be any point to be encoded in the second layer. The neighborhood point of the second point to be encoded is called the third neighborhood point. The embodiment of the present application does not specifically limit the definition method of the third neighborhood point. For example, the third neighborhood point may be a point in the same LOD layer as the second point to be encoded. Alternatively, the third neighborhood point may be a point in the next layer of the LOD layer where the second point to be encoded is located. Alternatively, the third neighborhood point may include points in the above two LOD layers at the same time. As an example, the third neighborhood point may be a point in the first layer (the upper layer of the LOD layer where the second point to be encoded is located) that is collinear with the second point to be encoded.

Step S2020 in Figure 20, i.e. predicting the shift coefficients of the points in the second layer, may include: predicting the second point to be encoded based on the shift coefficients of the third neighborhood points and the second prediction weight. In the related art, the prediction weight adopts a fixed value (i.e. 0.5). Different from the related art, in the embodiment of the present application, the second prediction weight is determined based on the weight corresponding to the third neighborhood point. For example, the greater the weight corresponding to a point in the third neighborhood point, the greater the influence of the point on the prediction result of the second point to be encoded. The prediction weight is determined based on the weight corresponding to the neighborhood point, which can make the points with higher importance have a greater influence on the prediction result of the point to be encoded, which will make the prediction process more reasonable and help improve the encoding efficiency of the geometric information of the points in the three-dimensional grid.

The weight corresponding to the third neighborhood point can be determined based on the spatial connection relationship of the third neighborhood point in the three-dimensional grid. For example, the weight corresponding to the third neighborhood point can be related to the number of neighboring points of the third neighborhood point (hereinafter referred to as the fourth neighborhood point). Determining the weight corresponding to the point based on the number of neighborhood points makes full use of the connection relationship information of the point in the three-dimensional grid, which can make the weight of the determined point more reasonable.

The embodiment of the present application does not specifically limit the definition of the fourth neighborhood point. For example, the fourth neighborhood point may be a point in the same LOD layer as the third neighborhood point. Alternatively, the fourth neighborhood point may be a point in the next layer of the LOD layer where the third neighborhood point is located. Alternatively, the fourth neighborhood point may include points in both LOD layers. As an example, the fourth neighborhood point is a point predicted based on the third neighborhood point.

In some implementations, the weight corresponding to the third neighborhood point may be equal to the number of the fourth neighborhood points. For example, the third neighborhood points include point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points, then the weight of point 1 may be 5, and the weight of point 2 may be 6.

In some implementations, the weight corresponding to the third neighborhood point can be determined based on the product of the number of fourth neighborhood points and the first prediction weight. For example, the weight corresponding to the third neighborhood point can be equal to the product of the number of fourth neighborhood points and the first prediction weight. The first prediction weight mentioned here can be a predefined fixed value. For example, the value of the first prediction weight can be 0.5. When determining the weight corresponding to the point, considering the prediction weight can make the determined weight more reasonable.

Exemplarily, the weight corresponding to the third neighborhood point may be determined based on the following formula:

In the above formula, weight represents the weight corresponding to any point in the third neighborhood point, and predWeight is the first prediction weight (the value can be 0.5). P represents the fourth neighborhood point, and weight[i] represents the initial weight corresponding to the i-th point in the fourth neighborhood point. The initial weight corresponding to the i-th point can be set to 1. For example, if the third neighborhood point includes point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points, then based on the above formula, the weight of point 1 can be 2.5, and the weight of point 2 can be 3 (the first prediction weight is 0.5).

In some implementations, the weight corresponding to the third neighboring point can be determined based on the weight corresponding to the LOD layer where the first neighboring point is located. For example, the weight corresponding to the third neighboring point can be equal to the weight corresponding to the LOD layer where the third neighboring point is located. The weight corresponding to the LOD layer can be understood as The level weight of a LOD layer can be determined based on the ratio of the number of points in the LOD layer to the total number of points to be decoded.

As mentioned above, the second prediction weight can be determined based on the weight corresponding to the third neighborhood point. For example, the second prediction weight can be equal to the weight corresponding to the third neighborhood point. Alternatively, the second prediction weight can be based on the product of the weight corresponding to the third neighborhood point and a certain parameter.

As mentioned above, the second point to be encoded can be predicted according to the shift coefficient of the third neighborhood point and the second prediction weight. A possible prediction method for the second point to be encoded is given below.

Exemplarily, the second to-be-encoded point may be predicted based on the following formula:

In the above formula, Signal represents the prediction result of the second point to be encoded, and Neigh ₁ and Neigh ₂ represent the shift coefficients of the two neighboring points collinear with the second point to be encoded. Neigh ₁ and Neigh ₂ are located in the upper layer of the LOD layer where the second point to be encoded is located. predWeight ₁ and predWeight ₂ represent the prediction weights corresponding to Neigh ₁ and Neigh ₂ (i.e., the second prediction weights mentioned above).

predWeight ₁ and predWeight ₂ can be calculated using the following formula:
predWeight ₁ = weight ₁
predWeight ₂ = weight ₂

sumWeight can be calculated based on the following formula:
sumWeight＝predWeight ₁ +predWeight ₂

The above describes the transformation operation and prediction operation of points in the three-dimensional grid in detail, and the following describes the quantization operation of points in the three-dimensional grid from the perspective of the third point to be encoded. The third point to be encoded can be any point to be encoded in the three-dimensional grid.

Step S2040 in Figure 20, i.e., quantizing the transformation coefficients of the points in the three-dimensional grid, may include: quantizing the transformation coefficients of the third point to be encoded according to the weight corresponding to the third point to be encoded. When quantizing the points to be encoded in the three-dimensional grid, the embodiment of the present application considers the weight corresponding to the point to be encoded. For example, a suitable quantization parameter may be selected according to the weight corresponding to the point to be encoded; or, the quantization result may be processed differently according to the weight corresponding to the point to be encoded. Considering the weight of the point to be encoded in the quantization process can make the quantization result more reasonable, which helps to improve the encoding efficiency of the geometric information of the points in the three-dimensional grid.

The weight corresponding to the third point to be encoded can be determined based on the spatial connection relationship of the third point to be encoded in the three-dimensional grid. For example, the weight corresponding to the third point to be encoded can be related to the number of neighboring points (hereinafter referred to as the fifth neighboring points) of the third point to be encoded. Determining the weight corresponding to the point based on the number of neighboring points makes full use of the connection relationship information of the point in the three-dimensional grid, which can make the weight of the determined point more reasonable.

The embodiment of the present application does not specifically limit the definition of the fifth neighborhood point. For example, the fifth neighborhood point may be a point in the same LOD layer as the third point to be encoded. Alternatively, the fifth neighborhood point may be a point in the next layer of the LOD layer where the third point to be encoded is located. Alternatively, the fifth neighborhood point may include points in both LOD layers. As an example, the fifth neighborhood point is a point predicted based on the third point to be encoded.

In some implementations, the weight corresponding to the third point to be encoded may be equal to the number of the fifth neighboring points. For example, the third point to be encoded includes point 1 and point 2, point 1 includes 5 neighboring points, and point 2 includes 6 neighboring points, then the weight of point 1 may be 5, and the weight of point 2 may be 6.

In some implementations, the weight corresponding to the third point to be encoded may be determined based on the product of the number of the fifth neighborhood points and the first prediction weight. For example, the weight corresponding to the third point to be encoded may be equal to the product of the number of the fifth neighborhood points and the first prediction weight. The first prediction weight mentioned here may be a predefined fixed value. For example, the value of the first prediction weight may be 0.5. When determining the weight corresponding to the point, considering the prediction weight may make the determined weight more reasonable.

Exemplarily, the weight corresponding to the third to-be-encoded point may be determined based on the following formula:

In the above formula, weight represents the weight corresponding to any point in the third point to be encoded, and predWeight is the first prediction weight (the value can be 0.5). P represents the fifth neighborhood point, and weight[i] represents the initial weight corresponding to the i-th point in the fifth neighborhood point. The initial weight corresponding to the i-th point can be set to 1. For example, the third point to be encoded includes point 1 and point 2, point 1 includes 5 neighborhood points, and point 2 includes 6 neighborhood points. Based on the above formula, the weight of point 1 can be 2.5, and the weight of point 2 can be 3 (the first prediction weight is 0.5).

As mentioned above, the transform coefficient of the third point to be coded may be quantized according to the weight corresponding to the third point to be coded. A possible implementation manner is given below.

For example, in some implementations, the transform coefficient of the third point to be encoded can be quantized according to the first quantization parameter. The first quantization parameter can be determined based on the weight corresponding to the third point to be encoded. In other words, the quantization parameter can be adaptively set for the point to be encoded according to the weight corresponding to the point to be encoded. For example, if the weight corresponding to the third point to be encoded is large, it means that the importance of the third point to be encoded is high. For points to be encoded with high importance, the value of the first quantization parameter can be small (indicating that the degree of quantization of the point to be encoded is low). For another example, if the weight corresponding to the third point to be encoded is small, it means that the importance of the third point to be encoded is low. For points to be encoded with low importance, the value of the first quantization parameter can be small (indicating that the degree of quantization of the point to be encoded is low). For code points, the value of the first quantization parameter can be relatively large (indicating that the quantization degree of the code point is relatively high). According to the design method of the quantization parameter, the reconstruction quality can be guaranteed, thereby improving the geometric coding efficiency of the points in the three-dimensional grid.

The first quantization parameter can be directly determined based on the weight corresponding to the third point to be encoded. For example, a mapping relationship between the weight corresponding to the point to be encoded and the quantization parameter can be pre-established, and the first quantization parameter can be determined based on the mapping relationship. Alternatively, the first quantization parameter can also be determined based on the weight corresponding to the third point to be encoded and the second quantization parameter. The second quantization parameter mentioned here can be the quantization parameter corresponding to the target LOD layer (i.e., the LOD layer to which the third point to be encoded belongs). The second quantization parameter can be determined based on the initial quantization parameter and the quantization adjustment parameter.

Exemplarily, the second quantization parameter may be determined based on the following formula:
QP _Lvl ＝liftingQP×Scale ^Lvl

In the above formula, Lvl represents the index of the target LOD layer, QP _Lvl is the second quantization parameter, liftingQP is the initial quantization parameter, and Scale is the quantization adjustment parameter.

A more specific example of quantization of the third point to be encoded is given below.

For example, the third to-be-encoded point may be quantized based on the following formula:

In the above formula, _QResi represents the quantization result of the third point to be encoded, Rea _i represents the transformation coefficient of the third point to be encoded, bitDepth represents the effective bit depth of the shift coefficient, QP _Lvl represents the quantization parameter corresponding to the target LOD layer (the LOD layer where the third point to be encoded is located), and weight _i represents the weight corresponding to the third point to be encoded (the method for determining the weight corresponding to the third point to be encoded can be found in the previous text and will not be described in detail here).

The above text, in conjunction with Figures 19 and 20, describes in detail the encoding and decoding method provided by the embodiment of the present application. It should be noted that, in the absence of conflict, the terms "coefficient" and "information" or "value" mentioned in any of the above embodiments can be used interchangeably. For example, the shift coefficient mentioned above can be replaced by a shift value or shift information, and the residual coefficient can be replaced by a residual value or residual information.

It should also be noted that in the above examples, the prediction operation corresponding to the decoding end can also be called an inverse prediction operation. The transform operation at the encoding end can also be called an update operation. The inverse transform operation at the decoding end can also be called an update operation or an inverse update operation.

The following is a more detailed description of the present application embodiments in conjunction with specific examples. It should be noted that the examples below are only intended to help those skilled in the art understand the present application embodiments, rather than limiting the present application embodiments to the illustrated specific numerical values or specific scenarios. Those skilled in the art can obviously make various equivalent modifications or changes based on the examples given, and such modifications or changes also fall within the scope of the present application embodiments.

Example 1:

Example 1 obtains the weight weight _i of each point based on the spatial connection relationship between adjacent points in the three-dimensional grid, where i = 0, 1, 2...pointCount, and pointCount represents the number of points in the three-dimensional grid. Then, the shift coefficient of each point is transformed with the help of the LOD spatial structure. Among them, the prediction scheme still adopts the prediction coding scheme provided by the relevant technology, that is, the shift coefficient of the middle point is predicted and encoded using the shift coefficients of two collinear adjacent points, and the prediction weight is fixed at 0.5. When updating the shift coefficient of each point, the update weight is adaptively obtained according to the weight weight _i of each point and the prediction weight.

1.1. Overall encoding process

As shown in Figure 21, according to the LOD spatial structure, the shift coefficients of the entire set of points to be encoded are divided into low-frequency coefficients L(N) and high-frequency coefficients H(N). Then, the low-frequency coefficients L(N) are used to predict the high-frequency coefficients H(N). Then, the predicted high-frequency coefficients H(N) are used to adaptively update the low-frequency coefficients L(N).

As shown in FIG22, a three-layer LOD spatial structure is shown, and the coding order is from the low LOD layer to the high LOD layer, that is, Lvl ₂ ->Lvl ₁ ->Lvl _0. First, the entire shift coefficient to be coded is divided into high-frequency coefficient Lvl ₂ and low-frequency coefficient Lvl ₀ +Lvl _1. After the coding of the shift coefficient of the Lvl ₂ layer is completed, the remaining shift coefficients are further divided into high-frequency coefficient Lvl ₁ and low-frequency coefficient Lvl ₀ .

1.2 Encoding algorithm

Assign an initial weight weight _i to all the points to be coded and set it to 1, that is, all the points to be coded are considered equally important.

The initial weights are updated according to the connection relationship between the points in the three-dimensional grid. The specific update formula is as follows:

Among them, predWeight is the prediction weight. The current V-DMC sets the prediction weight to a fixed value of 0.5. The P set represents the neighborhood point set predicted based on the current point.

Prediction is performed based on the LOD spatial structure. The specific prediction algorithm is shown in the following formula:
signal=predWeight×(Neigh ₁ +Neigh ₂ )

Among them, Neigh ₁ and Neigh ₂ represent the shift coefficients of two neighboring points that are collinear with the current point, and these two neighboring points are located in the upper LOD of the current layer.

After completing the prediction of the displacement coefficients of all points in the current layer, the displacement coefficients of the points in the upper LOD are updated using the predicted residual of the displacement coefficients of each point. The specific update formula is as follows:

The specific calculation method of updateWeight _i is as follows:
updateWeight _i = predWeight × weight _i

predWeight is the prediction weight, and weight _i is the weight of each point.

sumWeight is calculated as follows:

Signal _i represents the shifted residual coefficients of each point predicted based on the current point.

The above prediction and update process is repeated continuously, and the transformation is performed from the lower layer to the upper layer according to the spatial structure of the LOD, and finally the shift coefficient of the entire grid is transformed from the spatial domain to the frequency domain, thereby completing the geometric encoding of the entire grid.

1.3 Decoding algorithm

The decoding algorithm is a completely opposite process to the encoding algorithm. After completing the entropy decoding of the shift coefficients, the decoding end continuously performs inverse transformation from the high level of LOD to the lowest level of LOD according to the LOD spatial structure, thereby completing the geometric information reconstruction of the entire grid. The specific decoding algorithm is described as follows.

First, all the points to be decoded are assigned an initial weight weight _i set to 1, that is, all the points to be decoded are considered equally important.

The weights are updated according to the connection relationship between the points in the three-dimensional grid. The specific update formula is as follows:

Among them, predWeight is the prediction weight value. The current V-DMC sets the prediction weight to a fixed value of 0.5. The P set is the neighborhood point set predicted based on the current point.

According to the LOD spatial structure, the inverse transformation is performed, and the spatial connection relationship between points is used to obtain the neighborhood point set based on the current point for prediction. The specific inverse transformation formula is as follows:

The specific calculation method of updateWeight _i is as follows:
updateWeight _i = predWeight × weight _i

predWeight is the prediction weight, and weight _i is the weight of each point.

sumWeight is calculated as follows:

Signal _i represents the shifted residual coefficient of the point predicted based on the current point.

According to the spatial structure of LOD, reverse prediction is performed. The specific prediction algorithm is shown in the following formula:
signal+＝predWeight×(Neigh ₁ +Neigh ₂ )

Among them, Neigh ₁ and Neigh ₂ represent two neighboring points on the same line as the current point, and these two neighboring points are located in the upper LOD.

The inverse transformation and inverse prediction process described above is repeated continuously, and the inverse transformation is performed from the upper layer to the lower layer according to the LOD spatial structure, and finally the shift coefficients of the entire grid are transformed from the frequency domain to the spatial domain, thereby completing the geometric encoding of the entire grid.

After completing the inverse transformation of the shift coefficient of each point in the three-dimensional grid, the reconstructed geometric information of the three-dimensional grid can be determined by adding the initial geometric position information (geometric information of the subdivided grid) to the reconstructed shift coefficient of each point.

Example 1 proposes a shift coding scheme for lifting transformation based on the spatial connection relationship of the grid. Specifically, LOD space division is performed on the points after the basic grid division, and the shift of the points in the entire grid is lifted and transformed (including prediction and transformation) based on the LOD space structure. The prediction process uses two colinear neighboring points of the current point for prediction coding, and the transformation process adaptively updates the shift coefficient of the current point according to the shift coefficient, weight and prediction weight of each point in the set of neighboring points predicted based on the current point. In other words, Example 1 adjusts the fixed update weights used in the update algorithm to: update according to the number of neighboring points of each point to be encoded, the weight of each neighboring point, and the shift residual coefficient of the neighboring point. This coding scheme can further improve the geometric coding efficiency of the grid by effectively utilizing the spatial relationship between points to transform the shift coefficient of each point.

The technical solution provided in Example 1 was tested based on V-DMC-4.0 using test conditions C1 and C2. Taking the lossy test environment of geometric lossy attributes as an example, BD-Rate is a performance indicator for measuring compression efficiency. When BD-Rate is less than 0, it means that the encoding efficiency has been improved compared to the traditional encoding scheme. It can be seen from the table below that compared with V-DMC-4.0, under the condition of C1, D1 and D2 increased by approximately 2.4% and 2.4% respectively, and the Luma component increased by approximately 0.4%; under the condition of C2, D1 and D2 increased by 2.1% and 2.1% respectively, and the Luma component increased by approximately 0.4%.

Example 2:

Example 2 obtains the weight weight _i of each point based on the spatial connection relationship between adjacent points in the grid, where i = 0, 1, 2...pointCount, and pointCount represents the number of points in the three-dimensional grid. During the encoding process, first, the shift coefficient of the current grid is lifted and transformed to transform the shift coefficient of each point from the spatial domain to the frequency domain; then, after the transformation is completed, the shift coefficient is quantized. During the quantization process, the quantization parameter of the current point to be encoded can be adaptively determined according to the weight of the current point.

2.1 The entire encoding process

As shown in FIG22 , according to the LOD spatial structure, the points of the entire grid to be encoded are divided into three LOD layers. The encoding order is from the low LOD layer to the high LOD layer, that is, Lvl ₂ ->Lvl ₁ ->Lvl _0. First, the entire shift coefficient to be encoded is divided into high-frequency coefficients Lvl ₂ and low-frequency coefficients Lvl ₀ +Lvl _1. After completing the encoding of the shift coefficients of the Lvl ₂ layer, the remaining shift coefficients are further divided into high-frequency coefficients Lvl ₁ and low-frequency coefficients Lvl ₀ .

2.2 Encoding algorithm

Assign an initial weight weight _i to all the points to be coded and set it to 1, that is, all the points to be coded are considered to be equally important.

The initial weights are updated according to the connection relationship between the points in the three-dimensional grid. The specific update formula is as follows:

Among them, predWeight is the prediction weight. The current V-DMC sets the prediction weight to a fixed value of 0.5. The P set represents the neighborhood point set predicted based on the current point.

Prediction is performed based on the LOD spatial structure. The specific prediction algorithm is shown in the following formula:
Signal＝predWeight×(Neigh ₁ +Neigh ₂ )

Among them, Neigh ₁ and Neigh ₂ represent the shift coefficients of two neighboring points that are collinear with the current point, and these two neighboring points are located in the upper LOD of the current layer.

After completing the prediction of the displacement coefficients of all points in the current layer, the displacement coefficients of the midpoints in the upper LOD are updated using the prediction residual of each point displacement coefficient. The specific update formula is as follows:

Among them, updateWeight is the update weight, and P set is the point set predicted based on the current point.

The shift coefficient of the current layer is adaptively quantized. The specific quantization method is as follows:

Among them, _Resi represents the transformation coefficient of the shift coefficient of the current point, bitDepth represents the effective bit depth of the shift coefficient in the current geometric coding, QP _Lvl represents the quantization parameter of the current LOD layer, and weight _i is the weight corresponding to the current point.

According to the spatial structure of LOD, the prediction, update and quantization process described above is repeated to complete the geometric encoding of the entire grid.

2.3 Decoding algorithm

The decoding algorithm is a completely opposite process to the encoding algorithm. When the encoding end encodes the shift coefficient of the point, it continuously performs inverse transformation from the high level to the low level of the LOD according to the spatial structure of the LOD. After completing the entropy decoding of the shift coefficient, the decoding end continuously performs inverse transformation from the high level of the LOD to the low level of the LOD according to the spatial structure of the LOD, thereby completing the geometric information reconstruction of the entire mesh.

All the points to be decoded are assigned an initial weight weight _i set to 1, which means that all the points to be decoded are considered equally important.

The initial weights are updated according to the connection relationship between the points in the three-dimensional grid. The specific update formula is as follows:

Among them, predWeight is the prediction weight. The current V-DMC sets the prediction weight to a fixed value of 0.5. The P set represents the neighborhood point set predicted based on the current point.

The shift coefficients of the current layer are adaptively dequantized. The specific dequantization method is as follows:

Among them, Res _i represents the quantization coefficient of the shift coefficient of the current point, bitDepth represents the effective bit depth of the shift coefficient in the current geometric coding, QP _Lvl represents the quantization parameter of the current LOD layer, weight _i is the weight of the current point bet, and InvRes _i represents the inverse quantization result of the current point.

According to the spatial structure of LOD, the inverse transformation is performed. The specific inverse transformation formula is as follows:

Among them, updateWeight is the update weight, and P set is the set of neighborhood points predicted based on the current point.

According to the spatial structure of LOD, reverse prediction is performed. For specific prediction, see the following formula:
Signal＝predWeight×(Neigh ₁ +Neigh ₂ )

Among them, Neigh ₁ and Neigh ₂ represent the shift coefficients of two neighboring points on the same line as the current point, and these two neighboring points are located at the upper LOD layer of the current point.

The above-mentioned inverse quantization, inverse transformation and inverse prediction processes are continuously repeated to complete the geometric decoding of the entire grid.

After completing the inverse transformation of the shift coefficient of each point in the three-dimensional grid, the reconstructed geometric information of the three-dimensional grid can be determined by adding the initial geometric position information (geometric information of the subdivided grid) to the reconstructed shift coefficient of each point.

This scheme proposes a shift coding scheme for adaptive quantization based on the spatial connection relationship of the grid. Specifically, the points after the basic grid division are divided into LOD space; the shift of the points in the grid is lifted and transformed (including prediction and transformation) based on the spatial structure of LOD. In the prediction process, two colinear neighboring points of the current point are used for prediction coding. The transformation process will adaptively update the shift coefficient of the current point according to the shift coefficient of each point in the domain point set predicted based on the current point, and the fixed update weight. After completing the shift coefficient transformation, the shift coefficient is quantized and encoded. During quantization coding, the shift coefficients of different points are adaptively quantized and encoded according to the weights of different points in the current grid. The weight of the point can be obtained by utilizing the spatial connection relationship between the points in the current grid. For points with larger weights and higher importance, a smaller quantization degree can be used, and for points with smaller weights and lower importance, a higher quantization degree can be used. Based on such a scheme, the code rate of the geometric information of the grid can be reduced while ensuring the reconstruction quality of the entire grid, thereby further improving the coding efficiency of the geometric information of the grid.

The technical solution provided in Example 1 was tested based on V-DMC-4.0 using test conditions C1 and C2. Taking the lossy test environment of geometric lossy attributes as an example, BD-Rate is a performance indicator for measuring compression efficiency. When BD-Rate is less than 0, it means that the encoding efficiency has been improved compared to the traditional encoding scheme. It can be seen from the table below that compared with V-DMC-4.0, under the condition of C1, D1 and D2 are increased by approximately 0.2% and 0.2% respectively, and the Luma component is increased by approximately 0.0% to 0.1%; under the condition of C2, D1 and D2 are increased by 0.2% and 0.2% respectively, and the Luma component is increased by approximately 0.1%.

Example 3:

Based on Example 1 or Example 2, after obtaining the weight of the current point, the prediction weight of each point can be adaptively updated according to the weight of the current point, that is, the prediction algorithm of the current point is modified as follows:

Among them, predWeight ₁ can be obtained by using the weight of the first neighborhood point, that is,
predWeight ₁ = weight ₁
predWeight ₂ = weight ₂

Then, sumWeight is calculated as follows:
sumWeight＝ptedWeight ₁ +predWeight ₂

Based on such a coding scheme, when predictive coding is performed on the shift coefficient of the current point, the spatial connection relationship between the grid points is also used, thereby further improving the coding efficiency of the geometric information of the grid.

The method embodiment of the present application is described in detail above in conjunction with Figures 1 to 22, and the device embodiment of the present application is described in detail below in conjunction with Figures 23 to 26. It should be understood that the description of the method embodiment corresponds to the description of the device embodiment, so the part not described in detail can refer to the previous method embodiment.

FIG23 is a schematic diagram of the structure of a decoder provided by an embodiment of the present application. The decoder 2300 shown in FIG23 includes a first determination unit 2310, a second determination unit 2320, a third determination unit 2330, and a prediction unit 2340. The first determination unit 2310 is configured to parse the code stream and determine the quantization coefficients of the points in the three-dimensional grid. The second determination unit 2320 is configured to dequantize the quantization coefficients of the points in the three-dimensional grid and determine the dequantization coefficients of the points in the three-dimensional grid. The third determination unit 2330 is configured to perform an inverse transformation on the dequantization coefficients of the points in the first layer and determine the shift coefficients of the points in the first layer. The prediction unit 2340 is configured to predict the shift coefficients of the points in the second layer according to the shift coefficients of the points in the first layer. The points in the three-dimensional grid belong to multiple LOD layers, the first layer and the second layer are two adjacent layers in the multiple LOD layers, and the first layer is the upper layer of the second layer.

In some implementations, the third determination unit 2330 is configured to: perform an inverse transformation on the inverse quantization coefficient of the first point to be decoded in the first layer according to the inverse quantization coefficient of the first neighborhood point and the updated weight, wherein the first neighborhood point is a neighborhood point of the first point to be decoded, and the updated weight is determined based on the weight corresponding to the first neighborhood point.

In some implementations, the update weight is determined based on the weight corresponding to the first neighborhood point and the first prediction weight.

In some implementations, the updated weight is determined based on the product of the weight corresponding to the first neighborhood point and the first prediction weight.

In some implementations, the weight corresponding to the first neighborhood point is related to the number of second neighborhood points, and the second neighborhood points are neighborhood points of the first neighborhood point; or, the weight corresponding to the first neighborhood point is determined based on the weight corresponding to the LOD layer where the first neighborhood point is located.

In some implementations, the weight corresponding to the first neighborhood point is equal to the number of the second neighborhood points; or, the weight corresponding to the first neighborhood point is equal to the product of the number of the second neighborhood points and the first prediction weight.

In some implementations, the second neighborhood point is a point predicted based on the first neighborhood point.

In some implementations, the first neighborhood point is a point predicted based on the first point to be decoded.

In some implementations, the prediction unit 2340 is configured to: predict the second point to be decoded based on the shift coefficient of the third neighborhood point and the second prediction weight, the second point to be decoded is any point in the second layer, the third neighborhood point is the neighborhood point of the second point to be decoded in the first layer, and the second prediction weight is determined based on the weight corresponding to the third neighborhood point.

In some implementations, the second prediction weight is equal to the weight corresponding to the third neighborhood point.

In some implementations, the weight corresponding to the third neighborhood point is associated with the number of fourth neighborhood points, and the fourth neighborhood point is a neighborhood point of the third neighborhood point; or, the weight corresponding to the third neighborhood point is determined based on the weight corresponding to the LOD layer where the third neighborhood point is located.

In some implementations, the weight corresponding to the third neighborhood point is equal to the number of the fourth neighborhood points; or, the weight corresponding to the third neighborhood point is equal to the product of the number of the fourth neighborhood points and the first prediction weight.

In some implementations, the fourth neighborhood point is a point predicted based on the third neighborhood point.

In some implementations, the third neighborhood point is a point in the first layer that is colinear with the second point to be decoded.

In some implementations, the second determining unit 2320 is configured to: dequantize the quantization coefficient of the third point to be decoded according to a weight corresponding to the third point to be decoded, wherein the third point to be decoded is any point in the three-dimensional network.

In some implementations, the second determining unit 2320 is configured to: perform inverse quantization on the quantization coefficient of the third point to be decoded according to a first quantization parameter, where the first quantization parameter is determined based on a weight corresponding to the third point to be decoded.

In some implementations, the first quantization parameter is determined based on the weight corresponding to the third point to be decoded and the second quantization parameter, wherein the third point to be decoded is located at a target LOD layer among the multiple LOD layers, and the second quantization parameter is a quantization parameter corresponding to the target LOD layer.

In some implementations, the second quantization parameter is determined based on an initial quantization parameter and a quantization adjustment parameter.

In some implementations, the greater the weight corresponding to the third point to be decoded, the smaller the value of the first quantization parameter.

In some implementations, the weight corresponding to the third point to be decoded is related to the number of fifth neighborhood points, and the fifth neighborhood points are neighborhood points of the third point to be decoded.

In some implementations, the weight corresponding to the third point to be decoded is equal to the number of the fifth neighborhood points; or, the weight corresponding to the third point to be decoded is equal to the product of the number of the fifth neighborhood points and the first prediction weight.

In some implementations, the fifth neighborhood point is a point predicted based on the third point to be decoded.

In some implementations, the value of the first prediction weight is a predefined fixed value.

In some implementations, the value of the first prediction weight is 0.5.

In some implementations, the decoder 2300 also includes: a fourth determination unit, configured to: parse the code stream to determine the geometric information of the points in the basic grid; divide the basic grid according to the geometric information of the points in the basic grid to determine the geometric information of the points in the subdivided grid; determine the reconstructed geometric information of the points in the three-dimensional grid according to the shift coefficients of the points in the three-dimensional grid and the geometric information of the points in the subdivided grid.

It is understandable that in the embodiments of the present application, a "unit" may be a part of a circuit, a part of a processor, a part of a program or software, etc., and of course, it may be a module, or it may be non-modular. Moreover, the components in the present embodiment may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional module.

If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

Therefore, an embodiment of the present application provides a computer-readable storage medium, which is applied to the decoder 2300. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the decoding method described in any one of the aforementioned embodiments.

Based on the composition of the above-mentioned decoder and the computer-readable storage medium, refer to Figure 24, which shows a specific hardware structure diagram of the decoder provided in an embodiment of the present application. As shown in Figure 24, the decoder 2400 may include: a communication interface 2410, a memory 2420 and a processor 2430; each component is coupled together through a bus system 2440. It can be understood that the bus system 2440 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 2440 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, various buses are marked as bus system 2440 in Figure 24. Among them,

Communication interface 2410, used for receiving and sending signals during the process of sending and receiving information with other external network elements;

Memory 2420, used for storing computer programs;

The processor 2430 is configured to, when running the computer program, execute:

Parse the bitstream and determine the quantization coefficients of the points in the three-dimensional grid;

Dequantizing the quantized coefficients of the points in the three-dimensional grid to determine the dequantized coefficients of the points in the three-dimensional grid;

Performing an inverse transformation on the inverse quantization coefficients of the points in the first layer to determine the shift coefficients of the points in the first layer;

Predicting the shift coefficients of the points in the second layer according to the shift coefficients of the points in the first layer;

The points in the three-dimensional grid belong to multiple LOD layers, the first layer and the second layer are two adjacent layers in the multiple LOD layers, and the first layer is an upper layer of the second layer.

It can be understood that the memory 2420 in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DRRAM). The memory 2420 of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

The processor 2430 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 2430 or the instructions in software form. The above-mentioned processor 2430 can be a general-purpose processor, a digital signal processor (digital signal processor, DSP), an application specific integrated circuit (application specific integrated circuit, ASIC), a field programmable gate array (field programmable gate array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The various methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly reflected as being executed by a hardware decoding processor, or using a decoding processor. The hardware and software modules are combined and executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 2420, and the processor 2430 reads the information in the memory 2420 and completes the steps of the above method in combination with its hardware.

It is understood that the embodiments described in this application can be implemented in hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in this application or a combination thereof. For software implementation, the technology described in this application can be implemented by modules (such as processes, functions, etc.) that perform the functions described in this application. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

Optionally, as another embodiment, the processor 2430 is further configured to execute the decoding method described in any one of the aforementioned embodiments when running the computer program.

FIG25 is a schematic diagram of the structure of an encoder provided by an embodiment of the present application. The encoder 2500 of FIG25 includes a first determination unit 2510, a second determination unit 2520, and a transformation unit 2530. The first determination unit 2510 is configured to perform detail level LOD division on points in a three-dimensional grid, and determine multiple LOD layers, wherein the multiple LOD layers include adjacent first and second layers, and the first layer is the upper layer of the second layer. The second determination unit 2520 is configured to predict the shift coefficients of the points in the second layer and determine the residual coefficients of the points in the second layer. The transformation unit 2530 is configured to transform the shift coefficients of the points in the first layer according to the residual coefficients of the points in the second layer.

In some implementations, the transformation unit 2530 is configured to transform the shift coefficient of the first point to be encoded in the first layer according to the residual coefficient of the first neighborhood point and the update weight, wherein the first neighborhood point is the neighborhood point of the first point to be encoded in the second layer, and the update weight is determined based on the weight corresponding to the first neighborhood point.

In some implementations, the update weight is determined based on the weight corresponding to the first neighborhood point and the first prediction weight.

In some implementations, the updated weight is determined based on the product of the weight corresponding to the first neighborhood point and the first prediction weight.

In some implementations, the weight corresponding to the first neighborhood point is related to the number of second neighborhood points, and the second neighborhood points are neighborhood points of the first neighborhood point; or, the weight corresponding to the first neighborhood point is determined based on the weight corresponding to the LOD layer where the first neighborhood point is located.

In some implementations, the weight corresponding to the first neighborhood point is equal to the number of the second neighborhood points; or, the weight corresponding to the first neighborhood point is equal to the product of the number of the second neighborhood points and the first prediction weight.

In some implementations, the second neighborhood point is a point predicted based on the first neighborhood point.

In some implementations, the first neighborhood point is a point predicted based on the first point to be encoded.

In some implementations, the second determination unit 2520 is configured to: predict the second point to be encoded based on the shift coefficient of the third neighborhood point and the second prediction weight, the second point to be encoded is any point in the second layer, the third neighborhood point is a neighborhood point of the second point to be encoded, and the second prediction weight is determined based on the weight corresponding to the third neighborhood point.

In some implementations, the second prediction weight is equal to the weight corresponding to the third neighborhood point.

In some implementations, the weight corresponding to the third neighborhood point is associated with the number of fourth neighborhood points, and the fourth neighborhood point is a neighborhood point of the third neighborhood point; or, the weight corresponding to the third neighborhood point is determined based on the weight corresponding to the LOD layer where the third neighborhood point is located.

In some implementations, the weight corresponding to the third neighborhood point is equal to the number of the fourth neighborhood points; or, the weight corresponding to the third neighborhood point is equal to the product of the number of the fourth neighborhood points and the first prediction weight.

In some implementations, the fourth neighborhood point is a point predicted based on the third neighborhood point.

In some implementations, the third neighborhood point is a point in the first layer that is colinear with the second point to be encoded.

In some implementations, the encoder 2500 further includes a quantization unit configured to: quantize a transform coefficient of a third point to be encoded according to a weight corresponding to the third point to be encoded, wherein the third point to be encoded is any point in the three-dimensional network.

In some implementations, the quantization unit is configured to: quantize the transform coefficient of the third point to be encoded according to a first quantization parameter, and the first quantization parameter is determined based on a weight corresponding to the third point to be encoded.

In some implementations, the first quantization parameter is determined based on the weight corresponding to the third point to be encoded and the second quantization parameter, wherein the third point to be encoded is located in a target LOD layer among the multiple LOD layers, and the second quantization parameter is the quantization parameter corresponding to the target LOD layer.

In some implementations, the second quantization parameter is determined based on an initial quantization parameter and a quantization adjustment parameter.

In some implementations, the greater the weight corresponding to the third to-be-encoded point is, the smaller the value of the first quantization parameter is.

In some implementations, the weight corresponding to the third to-be-coded point is related to the number of fifth neighborhood points, and the fifth neighborhood points are The neighboring points of the third point to be encoded.

In some implementations, the weight corresponding to the third point to be encoded is equal to the number of the fifth neighborhood points; or, the weight corresponding to the third point to be encoded is equal to the product of the number of the fifth neighborhood points and the first prediction weight.

In some implementations, the fifth neighborhood point is a point predicted based on the third point to be encoded.

In some implementations, the value of the first prediction weight is a predefined fixed value.

In some implementations, the value of the first prediction weight is 0.5.

In some implementations, the encoder 2500 also includes a third determination unit, configured to divide the basic grid according to the geometric information of the points in the basic grid, and determine the geometric information of the points in the subdivided grid; and determine the shift coefficients of the points in the three-dimensional grid according to the geometric information of the points in the subdivided grid and the original geometric information of the three-dimensional grid.

It is understandable that in the embodiments of the present application, a "unit" may be a part of a circuit, a part of a processor, a part of a program or software, etc., and of course, it may be a module, or it may be non-modular. Moreover, the components in the present embodiment may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional module.

If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

Therefore, an embodiment of the present application provides a computer-readable storage medium, which is applied to the encoder 2500. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the decoding method described in any one of the aforementioned embodiments.

Based on the composition of the above-mentioned encoder and the computer-readable storage medium, refer to Figure 26, which shows a specific hardware structure diagram of the encoder 2600 provided in an embodiment of the present application. As shown in Figure 26, the encoder 2600 may include: a communication interface 2610, a memory 2620 and a processor 2630; each component is coupled together through a bus system 2640. It can be understood that the bus system 2640 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 2640 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, various buses are marked as bus system 2640 in Figure 26. Among them,

Communication interface 2610, used for receiving and sending signals during the process of sending and receiving information with other external network elements;

Memory 2620, used for storing computer programs;

The processor 2630 is configured to, when running the computer program, execute:

Performing LOD division on points in the three-dimensional grid to determine a plurality of LOD layers, wherein the plurality of LOD layers include a first layer and a second layer adjacent to each other, and the first layer is a layer above the second layer;

Predicting the shift coefficients of the points in the second layer to determine the residual coefficients of the points in the second layer;

The shift coefficients of the points in the first layer are transformed according to the residual coefficients of the points in the second layer.

It is understood that the memory 2620 in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a ROM, a PROM, an EPROM, an EEPROM or a flash memory. The volatile memory can be a RAM, which is used as an external cache. By way of exemplary but not limiting explanation, many forms of RAM are available, such as SRAM, DRAM, SDRAM, DDRSDRAM, ESDRAM, SLDRAM and DRRAM. The memory 2620 of the system and method described in the present application is intended to include, but is not limited to, these and any other suitable types of memories.

The processor 2630 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of the hardware in the processor 2630 or an instruction in the form of software. The above-mentioned processor 2630 may be a general-purpose processor, DSP, ASIC, FPGA or other programmable logic device, discrete gate or transistor logic device, discrete hardware component. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor to be executed, or a combination of hardware and software modules in the decoding processor to be executed. The software module may be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 2620, and the processor 2630 reads the information in the memory 2620 and completes the steps of the above method in combination with its hardware.

It is to be understood that these embodiments described in the present application can be implemented with hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more ASICs, DSPs, DSPDs, PLDs, FPGAs, general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application or a combination thereof. For software implementation, the technology described in the present application can be realized by the module (e.g., process, function, etc.) that performs the functions described in the present application. The software code can be stored in a memory and executed by a processor. The memory can be implemented in a processor or outside a processor.

Optionally, as another embodiment, the processor 2630 is further configured to execute the encoding method described in any one of the aforementioned embodiments when running the computer program.

An embodiment of the present application also provides a computer-readable storage medium, which is a non-volatile computer-readable storage medium for storing a bit stream. The bit stream can be generated by an encoding method using an encoder, or the bit stream can be decoded by a decoding method using a decoder, wherein the decoding method can be the decoding method described in any of the preceding embodiments, and the encoding method can be the encoding method described in any of the preceding embodiments.

It should be noted that, in this application, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

The methods disclosed in several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

The features disclosed in several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

The features disclosed in several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims (56)

A decoding method, applied to a decoder, comprising: Parse the bitstream and determine the quantization coefficients of the points in the three-dimensional grid; Dequantizing the quantized coefficients of the points in the three-dimensional grid to determine the dequantized coefficients of the points in the three-dimensional grid; Performing an inverse transformation on the inverse quantization coefficients of the points in the first layer to determine the shift coefficients of the points in the first layer; Predicting the shift coefficients of the points in the second layer according to the shift coefficients of the points in the first layer; The points in the three-dimensional grid belong to multiple detail level LOD layers, the first layer and the second layer are two adjacent layers in the multiple LOD layers, and the first layer is an upper layer of the second layer. The method according to claim 1, wherein the inverse transformation of the inverse quantized coefficients of the points in the first layer comprises: According to the inverse quantization coefficient of the first neighborhood point and the updated weight, the inverse transformation is performed on the inverse quantization coefficient of the first point to be decoded in the first layer, wherein the first neighborhood point is a neighborhood point of the first point to be decoded, and the updated weight is determined based on the weight corresponding to the first neighborhood point. The method according to claim 2, wherein the update weight is determined based on the weight corresponding to the first neighborhood point and the first prediction weight. The method according to claim 3, wherein the updated weight is determined based on the product of the weight corresponding to the first neighborhood point and the first prediction weight. The method according to any one of claims 2 to 4, wherein the weight corresponding to the first neighborhood point is related to the number of second neighborhood points, and the second neighborhood points are neighborhood points of the first neighborhood points; or, the weight corresponding to the first neighborhood point is determined based on the weight corresponding to the LOD layer where the first neighborhood point is located. The method according to claim 5, wherein: The weight corresponding to the first neighborhood point is equal to the number of the second neighborhood points; or, The weight corresponding to the first neighborhood point is equal to the product of the number of the second neighborhood points and the first prediction weight. The method according to claim 5 or 6, wherein the second neighborhood point is a point predicted based on the first neighborhood point. The method according to any one of claims 2 to 7, wherein the first neighborhood point is a point predicted based on the first point to be decoded. The method according to any one of claims 1 to 8, wherein predicting the shift coefficients of the points in the second layer according to the shift coefficients of the points in the first layer comprises: The second point to be decoded is predicted according to the shift coefficient of the third neighborhood point and the second prediction weight, wherein the second point to be decoded is any point in the second layer, the third neighborhood point is a neighborhood point of the second point to be decoded in the first layer, and the second prediction weight is determined based on the weight corresponding to the third neighborhood point. The method according to claim 9, wherein the second prediction weight is equal to the weight corresponding to the third neighborhood point. The method according to claim 9 or 10, wherein the weight corresponding to the third neighborhood point is associated with the number of fourth neighborhood points, and the fourth neighborhood point is a neighborhood point of the third neighborhood point; or, the weight corresponding to the third neighborhood point is determined based on the weight corresponding to the LOD layer where the third neighborhood point is located. The method according to claim 11, wherein: The weight corresponding to the third neighborhood point is equal to the number of the fourth neighborhood points; or, The weight corresponding to the third neighborhood point is equal to the product of the number of the fourth neighborhood points and the first prediction weight. The method according to claim 11 or 12, wherein the fourth neighborhood point is a point predicted based on the third neighborhood point. The method according to any one of claims 9 to 13, wherein the third neighborhood point is a point in the first layer that is collinear with the second point to be decoded. The method according to any one of claims 1 to 14, wherein the dequantizing of the quantized coefficients of the points in the three-dimensional grid comprises: The quantization coefficient of the third point to be decoded is inversely quantized according to the weight corresponding to the third point to be decoded, wherein the third point to be decoded is any point in the three-dimensional network. The method according to claim 15, wherein the dequantizing the quantized coefficient of the third point to be decoded according to the weight corresponding to the third point to be decoded comprises: The quantization coefficient of the third point to be decoded is inversely quantized according to a first quantization parameter, where the first quantization parameter is determined based on a weight corresponding to the third point to be decoded. The method according to claim 16, wherein the first quantization parameter is determined based on the weight corresponding to the third point to be decoded and the second quantization parameter, wherein the third point to be decoded is located at a target LOD layer among the multiple LOD layers, and the second quantization parameter is a quantization parameter corresponding to the target LOD layer. The method according to claim 17, wherein the second quantization parameter is determined based on an initial quantization parameter and a quantization adjustment parameter. The method according to any one of claims 16 to 18, wherein the greater the weight corresponding to the third point to be decoded, the smaller the value of the first quantization parameter. The method according to any one of claims 15 to 19, wherein the weight corresponding to the third point to be decoded is related to the number of fifth neighborhood points, and the fifth neighborhood points are neighborhood points of the third point to be decoded. The method according to claim 20, wherein: The weight corresponding to the third to-be-decoded point is equal to the number of the fifth neighborhood points; or, The weight corresponding to the third to-be-decoded point is equal to the product of the number of the fifth neighborhood points and the first prediction weight. The method according to claim 20 or 21, wherein the fifth neighborhood point is a point predicted based on the third point to be decoded. The method according to claim 3, 4, 6, 12 or 21, wherein the value of the first prediction weight is a predefined fixed value. The method according to claim 23, wherein the value of the first prediction weight is 0.5. The method according to any one of claims 1 to 24, wherein the method further comprises: Parse the code stream to determine the geometric information of the points in the basic grid; Dividing the basic grid according to the geometric information of the points in the basic grid to determine the geometric information of the points in the subdivided grid; Reconstructed geometric information of the points in the three-dimensional grid is determined according to the shift coefficients of the points in the three-dimensional grid and the geometric information of the points in the subdivided grid. A coding method, applied to an encoder, comprising: Performing level of detail (LOD) division on points in the three-dimensional grid to determine a plurality of LOD layers, wherein the plurality of LOD layers include a first layer and a second layer adjacent to each other, and the first layer is a layer above the second layer; Predicting the shift coefficients of the points in the second layer to determine the residual coefficients of the points in the second layer; The shift coefficients of the points in the first layer are transformed according to the residual coefficients of the points in the second layer. The method according to claim 26, wherein transforming the shift coefficients of the points in the first layer according to the residual coefficients of the points in the second layer comprises: The shift coefficient of the first point to be encoded in the first layer is transformed according to the residual coefficient of the first neighborhood point and the update weight, wherein the first neighborhood point is the neighborhood point of the first point to be encoded in the second layer, and the update weight is determined based on the weight corresponding to the first neighborhood point. The method according to claim 27, wherein the updated weight is determined based on the weight corresponding to the first neighborhood point and the first prediction weight. The method according to claim 28, wherein the updated weight is determined based on the product of the weight corresponding to the first neighborhood point and the first prediction weight. The method according to any one of claims 27 to 29, wherein the weight corresponding to the first neighborhood point is related to the number of second neighborhood points, and the second neighborhood points are neighborhood points of the first neighborhood points; or, the weight corresponding to the first neighborhood point is determined based on the weight corresponding to the LOD layer where the first neighborhood point is located. The method according to claim 30, wherein: The weight corresponding to the first neighborhood point is equal to the number of the second neighborhood points; or, The weight corresponding to the first neighborhood point is equal to the product of the number of the second neighborhood points and the first prediction weight. The method according to claim 30 or 31, wherein the second neighborhood point is a point predicted based on the first neighborhood point. The method according to any one of claims 27 to 32, wherein the first neighborhood point is a point predicted based on the first point to be encoded. The method according to any one of claims 26 to 33, wherein predicting the shift coefficients of the points in the second layer comprises: The second point to be encoded is predicted according to the shift coefficient of the third neighborhood point and the second prediction weight, where the second point to be encoded is any point in the second layer, the third neighborhood point is a neighborhood point of the second point to be encoded, and the second prediction weight is determined based on the weight corresponding to the third neighborhood point. The method according to claim 34, wherein the second prediction weight is equal to the weight corresponding to the third neighborhood point. The method according to claim 34 or 35, wherein the weight corresponding to the third neighborhood point is associated with the number of fourth neighborhood points, and the fourth neighborhood point is a neighborhood point of the third neighborhood point; or the weight corresponding to the third neighborhood point is based on the third neighborhood point. The weight corresponding to the LOD layer where the point is located is determined. The method according to claim 36, wherein: The weight corresponding to the third neighborhood point is equal to the number of the fourth neighborhood points; or, The weight corresponding to the third neighborhood point is equal to the product of the number of the fourth neighborhood points and the first prediction weight. The method according to claim 36 or 37, wherein the fourth neighborhood point is a point predicted based on the third neighborhood point. The method according to any one of claims 34 to 38, wherein the third neighborhood point is a point in the first layer that is collinear with the second point to be encoded. The method according to any one of claims 26 to 39, wherein the method further comprises: The transformation coefficient of the third point to be encoded is quantized according to the weight corresponding to the third point to be encoded, wherein the third point to be encoded is any point in the three-dimensional network. The method according to claim 40, wherein the quantizing the transform coefficient of the third point to be coded according to the weight corresponding to the third point to be coded comprises: The transform coefficient of the third point to be encoded is quantized according to a first quantization parameter, where the first quantization parameter is determined based on a weight corresponding to the third point to be encoded. The method according to claim 41, wherein the first quantization parameter is determined based on the weight corresponding to the third point to be encoded and the second quantization parameter, wherein the third point to be encoded is located in a target LOD layer among the multiple LOD layers, and the second quantization parameter is the quantization parameter corresponding to the target LOD layer. The method of claim 42, wherein the second quantization parameter is determined based on an initial quantization parameter and a quantization adjustment parameter. According to the method according to any one of claims 41 to 43, wherein the greater the weight corresponding to the third point to be encoded, the smaller the value of the first quantization parameter. The method according to any one of claims 40 to 44, wherein the weight corresponding to the third point to be encoded is related to the number of fifth neighborhood points, and the fifth neighborhood points are neighborhood points of the third point to be encoded. The method of claim 45, wherein: The weight corresponding to the third to-be-encoded point is equal to the number of the fifth neighborhood points; or, The weight corresponding to the third point to be encoded is equal to the product of the number of the fifth neighborhood points and the first prediction weight. The method according to claim 45 or 46, wherein the fifth neighborhood point is a point predicted based on the third point to be encoded. The method according to claim 28, 29, 31, 37 or 46, wherein the value of the first prediction weight is a predefined fixed value. The method according to claim 48, wherein the value of the first prediction weight is 0.5. The method according to any one of claims 26 to 49, wherein the method further comprises: Dividing the basic grid according to the geometric information of the points in the basic grid, and determining the geometric information of the points in the subdivided grid; The shift coefficients of the points in the three-dimensional grid are determined according to the geometric information of the points in the subdivided grid and the original geometric information of the three-dimensional grid. A decoder, comprising: A first determination unit is configured to parse the bit stream and determine the quantization coefficients of the points in the three-dimensional grid; A second determining unit is configured to dequantize the quantization coefficients of the points in the three-dimensional grid to determine the dequantization coefficients of the points in the three-dimensional grid; a third determining unit, configured to perform an inverse transformation on the inverse quantization coefficients of the points in the first layer to determine the shift coefficients of the points in the first layer; a prediction unit configured to predict shift coefficients of points in the second layer according to the shift coefficients of points in the first layer; The points in the three-dimensional grid belong to multiple detail level LOD layers, the first layer and the second layer are two adjacent layers in the multiple LOD layers, and the first layer is an upper layer of the second layer. A decoder, comprising: Memory for storing computer programs; A processor, configured to execute the method according to any one of claims 1 to 25 when running the computer program. An encoder, comprising: A first determining unit is configured to perform detail level LOD division on points in the three-dimensional grid to determine a plurality of LOD layers, wherein the plurality of LOD layers include a first layer and a second layer adjacent to each other, and the first layer is a layer above the second layer; The second determining unit is configured to predict the shift coefficients of the points in the second layer and determine the residual coefficients of the points in the second layer. number; The transform unit is configured to transform the shift coefficients of the points in the first layer according to the residual coefficients of the points in the second layer. An encoder, comprising: Memory for storing computer programs; A processor, configured to execute the method according to any one of claims 16 to 50 when running the computer program. A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 25 or the method according to any one of claims 26 to 50 is implemented. A non-volatile computer-readable storage medium storing a bit stream, wherein the bit stream is generated by an encoding method using an encoder, or the bit stream is decoded by a decoding method using a decoder, wherein the decoding method is the method described in any one of claims 1 to 25, and the encoding method is the method described in any one of claims 26 to 50.