KR20080007069A - Video signal decoding method and apparatus - Google Patents

Abstract

The present invention relates to a video signal decoding method and apparatus. In the video signal decoding method using inter-view prediction, the method comprises: reading a macroblock type of a current macroblock and extracting macroblock information of a neighboring picture according to the read macroblock type; And decoding the current macroblock by using. Accordingly, the method and apparatus for decoding a video signal according to the present invention are capable of various prediction methods capable of removing redundant information between viewpoints in multi-view video coding, and can be used to remove duplicate information between viewpoints. By using the parity map, the coding efficiency is increased.

Description

Method and apparatus for decoding video signal {Method and Apparatus for decoding a video signal}

1 is a diagram illustrating a multiview image encoding and decoding system according to the present invention.

2 is a diagram illustrating a prediction structure between pictures in multi-view video coding according to the present invention.

3 (a) and 3 (b) are flowcharts illustrating an embodiment of a video signal decoding method according to the present invention.

4 to 9 are diagrams illustrating an embodiment of a video signal decoding method of FIGS. 3A and 3B.

10 is a flowchart illustrating another embodiment of a video signal decoding method according to the present invention.

11 is a block diagram showing an embodiment of a video signal decoding apparatus according to the present invention.

12 (a) and 12 (b) are flowcharts illustrating still another embodiment of a video signal decoding method according to the present invention.

FIG. 13 is a diagram illustrating an embodiment of a video signal decoding method of FIG. 12.

The present invention relates to a video signal decoding method and apparatus, and more particularly, to a video signal decoding method and apparatus by inter-view prediction.

The mainstream video broadcasting image is a single view image acquired with one camera. Multi-view video, on the other hand, is a three-dimensional (3D) image processing method that geometrically corrects images taken by more than one camera and provides users with various viewpoints in various directions through spatial synthesis. It is a field. Multi-view video can increase the freedom of view for the user, and has a feature that includes a larger area than the image area that can be acquired using a single camera.

Since such multi-view video images have a high correlation between viewpoints, redundant information can be removed through spatial prediction between viewpoints. Therefore, various prediction methods for prediction between viewpoints are needed.

An object of the present invention is to provide a video signal decoding method and apparatus for providing various prediction methods for removing redundant information between viewpoints.

Another object of the present invention is to provide a video signal decoding method using a disparity map to remove redundant information between viewpoints.

In order to achieve the above object, the video signal decoding method according to the present invention relates to a video signal decoding method using inter-view prediction, the method comprising the steps of: reading a macroblock type of a current macroblock; Decoding the current macroblock using macroblock information of a neighboring picture according to a macroblock type.

In addition, in order to achieve the above object, the video signal decoding method according to the present invention relates to a video signal decoding method using inter-view prediction, the current macroblock is a macro by intra-view prediction or inter-view prediction Reading a prediction mode indicating whether the block is a block; and decoding the current macroblock using macroblock information of a neighboring picture according to the read prediction mode.

In addition, to achieve the above object, the video signal decoding device according to the present invention relates to a video signal decoding apparatus using inter-view prediction, the macroblock for reading the macroblock type of the current macroblock And a type reading unit, a prediction method information reading unit for reading the prediction method information of the current macroblock, and a decoding unit decoding the current macroblock using the macroblock type and the prediction method information.

In order to achieve the above object, the video signal decoding method according to the present invention relates to a video signal decoding method using inter-view prediction, the method comprising: obtaining a pixel-by-pixel disparity vector; Deriving a block-by-block disparity vector from the parity vector and decoding the current macroblock using a disparity vector, which is a combination of the blocks having the same disparity vector among the derived block-based disparity vectors.

Hereinafter, a configuration and an operation of an embodiment of a video signal decoding apparatus according to the present invention and an embodiment of a method performed in the apparatus will be described with reference to the accompanying drawings. At this time, the configuration and operation of the present invention shown in the drawings and described by it will be described as at least one embodiment, by which the technical spirit of the present invention and its core configuration and operation is not limited.

1 is a diagram illustrating a multi-view sequence encoding and decoding system according to the present invention.

As shown in FIG. 1, the multi-view image encoding system to which the present invention is applied includes a multi-view image generator 10, a preprocessing unit 20, and an encoding unit 30. In addition, the decoding system includes an extractor 40, an extractor 50, a decoding unit, a post-processing unit 60, and a display unit 70.

The multi-view image generator 10 is provided with an image obtaining apparatus (for example, cameras # 1 to #N) corresponding to the number of multi-views to acquire independent images for each viewpoint. When the multiview image data is input, the preprocessing unit 20 performs a function of increasing the correlation between the multiview image data through a preprocessing process while solving noise removal and imbalancing problems. The encoder 30 includes a motion estimator, a motion compensator, and a disparity estimator, a disparity compensator, an illumination compensator, a bit rate control, and a residual image encoder. do. In addition, the encoding unit 30 may apply a generally known method.

The extractor 40 functions to extract only a bitstream corresponding to a desired view from the transmitted MVC bitstream. The extractor can view the header and optionally decode only the desired time point. In addition, view scalability may be implemented by extracting only a bitstream corresponding to a desired view using a view identifier that distinguishes a view of a picture. Since MVC must be fully compatible with H.264 / AVC, it is necessary to decode only certain views that are compatible with H.264 / AVC. In this case, a view identifier that distinguishes a view of the picture may be used to decode only compatible views. The bitstream extracted through the extractor 40 is transmitted to the decoding unit 50. The decoder 50 includes a motion compensator, an illumination compensator, a weight predictor, a deblocking filter, and the like. The decoding unit 50 receives the bitstream encoded by the above-described method, and decodes it in reverse. This will be described in detail with reference to FIG. 11. In addition, the post-processing unit 60 performs a function of increasing the reliability and resolution of the decoded data. Finally, the display unit 70 provides the decoded data to the user in various ways according to the function of the display, in particular, the ability to process a multi-view image. For example, it is a 2D display 71 that provides only a planar two-dimensional image, or a stereo type display 73 that provides two views as a stereoscopic image, or a stereoscopic image of M views (2 <M). The display 75 may be provided.

2 is a diagram illustrating a prediction structure between pictures in multi-view video coding according to the present invention.

As illustrated in FIG. 2, T0 to T100 on the horizontal axis represent frames according to time, and S0 to S100 on the vertical axis represent frames according to viewpoints, respectively. For example, pictures in T0 refer to images taken by different cameras in the same time zone (T0), and pictures in S0 refer to images in different time zones taken by one camera. In addition, the arrows in the drawing indicate the prediction direction and the order of each picture. For example, a P0 picture at S2 time point in the T0 time zone is a picture predicted from I0, and the P0 picture is also at S4 time point in the TO time zone. It becomes a reference picture of another P0 picture. It is also a reference picture of the B1 and B2 pictures in the T4 and T2 time zones at the time S2.

As such, when prediction encoding is performed using pictures that are in the same time zone as the current picture but at different views, an identifier for distinguishing the viewpoints between the pictures is required. Therefore, an identifier indicating a viewpoint of an arbitrary picture is defined as a viewpoint identifier (view_num). For example, the P0 picture at S4 in the TO time zone is predictively encoded by referring to the P0 picture at S2 in the same time zone. In this case, when referring to the P0 picture of the S2 view, more efficient predictive encoding can be performed by using the view identifier of the P0 picture of the S2 view.

3 (a) and 3 (b) are flowcharts illustrating an embodiment of a video signal decoding method according to the present invention.

As shown in FIG. 3 (a), first, the macroblock type of the current macroblock is read (S32). Here, according to the present invention, the macroblock type indicates whether the current macroblock type is a macroblock by intra-view prediction or a macroblock by inter-view prediction.

Next, the prediction method information of the current macroblock is read (S34). Here, the prediction method information according to the present invention may be divided into three pieces of information on motion prediction and two pieces of information on texture prediction. Information about motion prediction is described in detail with reference to FIG. 4 and information about texture prediction is described with reference to FIG. 7.

Finally, the current macroblock is decoded according to the read macroblock type and prediction method information (S36). In the video signal decoding method according to the present invention, macroblock information of a picture neighboring in a time direction or a view direction of a current macroblock is determined according to whether the read macroblock type is an intra-view prediction macroblock or an inter-view prediction macroblock. To decode the current macroblock. Here, the macroblock information of the neighboring picture may be divided into motion information and texture information, which will be described in detail below. Also, according to the read prediction information, macroblocks of pictures neighboring in the time direction or the view direction are macroblocks of pictures forward in the time direction or the view direction, or backward in the time direction or the view direction. This is the macroblock of the picture in. In other words, the macroblock type determines whether a picture neighboring the current macroblock is a picture neighboring in the time direction or a picture neighboring in the view direction. It can be seen whether the picture is in the forward or the picture in the backward.

3B, first, the macroblock type of the current macroblock is read out (S31). Here, according to the present invention, the macroblock type indicates whether the current macroblock type is a macroblock by intra-view prediction or a macroblock by inter-view prediction.

Next, prediction flag information indicating whether to apply the prediction method is read (S33). Here, the prediction flag information according to the present invention can be divided into three pieces of information on motion prediction and two pieces of information on texture prediction. Information about motion prediction will be described in detail with reference to FIGS. 5 and 6, and information about texture prediction will be described in detail with reference to FIGS. 8 and 9.

Finally, the current macroblock is decoded according to the read macroblock type and the prediction flag information (S35). In the video signal decoding method according to the present invention, macroblock information of a picture neighboring in a time direction or a view direction of a current macroblock is determined according to whether the read macroblock type is an intra-view prediction macroblock or an inter-view prediction macroblock. To decode the current macroblock. Here, the macroblock information of the neighboring picture may be divided into motion information and texture information, which will be described in detail below. Further, according to the read prediction flag information, macroblocks of pictures neighboring in the time direction or the view direction are macros of pictures forward in the time direction or the view direction when the prediction method is the forward prediction method. If the prediction method is a backward prediction method, the block is a macroblock of a picture backward in the time direction or the view direction.

4 to 9 are diagrams illustrating an embodiment of a video signal decoding method of FIGS. 3A and 3B.

4 (a), 5 (a), 6 (a), 7 (a), 8 (a), and 9 (a), the macroblocks of neighboring pictures in the video signal decoding method according to the present invention are selected from the current macroblocks. This is the case of a macroblock of a position corresponding to. In addition, it is assumed that the coordinate of the current macroblock is (x, y) and the coordinate of the macroblock at the corresponding position is (x + h, y + v). Here, the macroblocks MB_a, MB_b, MB_c, and MB_d at positions corresponding to the current macroblocks may be motion vectors (mv_l0, mv_l1) or disparity vectors (dis_l0, dis_l1) of the current macroblocks. Can be found using

4 (b), 5 (b), 6 (b), 7 (b), 8 (b) and 9 (b) show that macroblocks of neighboring pictures among video signal decoding methods according to the present invention are presently present. The case of the macroblock at the same position as the macroblock is shown.

FIG. 4 is a diagram for describing the video signal decoding method of FIG. 3 (a) in detail and relates to motion prediction among prediction scheme information in FIG. 3 (a).

First, the motion skip mode among the prediction method information will be described. In the present invention, the motion skip mode refers to decoding the current macroblock by using the motion or disparity vector and the reference index of the macroblock in the neighboring picture as the motion or disparity vector and the reference index of the current macroblock. To this end, the present invention newly defines a syntax called 'motion_skip_mode'.

For example, when the value of 'motion_skip_mode' is '0', the macroblock type of the current macroblock is not used. That is, the motion skip mode according to the present invention is not used. In other words, decoding is performed by a general video coding method. When the value of 'motion_skip_mode' is '1', it means forward prediction, and the macroblocks MB_a and MB_c of the picture forward in the time direction or the view direction of the picture to which the current macroblock belongs. Motion or disparity vector and reference index are used as the motion information of the current macroblock. When the value of 'motion_skip_mode' is '2', it means backward prediction, and the macroblock (MB_b) of a picture that is backward in the time direction or the view direction of the picture to which the current macroblock belongs The motion or disparity vector of MB_d) and the reference index are used as motion information of the current macroblock.

 This will be described with reference to FIGS. 4 (a) and 4 (b). First, it is assumed that the macroblock type of the current macroblock is an intra prediction macroblock, and the value of 'motion_skip_mode' is '1'.

Accordingly, the macroblocks of neighboring pictures become macroblocks MB_a and MB_b of neighboring pictures in the view direction by the macroblock type, and the macroblocks of the pictures previously in the view direction by the value of 'motion_skip_mode'. The motion vector and the reference index of are used as the motion vector and the reference index of the current macroblock.

In other words, in Fig. 4 (a), 'mv_l0 (x, y) = mv_a_l0 (x + h, y + v)' and 'mv_l1 (x, y) = mv_a_l1 (x + h, y + v)' The reference index of MB_a becomes the reference index of the current macroblock. In FIG. 4 (b), 'mv_t_0 = mv_ta_0' and 'mv_t_1 = mv_ta_1', and the reference index of MB_a becomes the reference index of the current macroblock.

Next, the motion refinement mode will be described. In the present invention, the motion refinement mode refers to the sum of the motion or disparity vector of a macroblock in a neighboring picture and the quarter-sample refinements of the current macroblock. In this case, the current macroblock is decoded using the reference index of the macroblock in the neighboring picture as the reference index of the current macroblock. That is, to use the motion or disparity vector of the neighboring picture as a predictor of the motion or disparity vector for the current macroblock. To this end, in the present invention, a syntax 'mv_refinement_mode' is newly defined.

For example, when the value of 'mv_refinement_mode' is '0', the macroblock type of the current macroblock is not used. In other words, the motion refinement mode according to the present invention is not used. When the value of 'mv_refinement_mode' is '1', it means forward prediction, and the macroblocks MB_a and MB_c of the picture forward in the time direction or the view direction of the picture to which the current macroblock belongs. The sum of the motion or disparity vector of) and the quarter-sample refinements of the current macroblock is used as the motion or disparity vector of the current macroblock. Herein, the present invention newly defines syntaxes 'mvd_refinement_l0' and 'mvd_refinement_l1' for quarter-sample refinement of the current macroblock. The reference index is the same as the motion skip mode. When the value of 'motion_skip_mode' is '2', it means backward prediction, and the macroblock (MB_b) of a picture that is backward in the time direction or the view direction of the picture to which the current macroblock belongs The sum of the motion or disparity vector of MB_d, the motion or disparity vector of the current macroblocks MB_a, MB_c, and the quarter-sample refinements of the current macroblock, is added to the motion or disparity of the current macroblock. Used as a parity vector. The reference index is the same as the motion skip mode.

This will be described with reference to FIGS. 4A and 4B. First, it is assumed that a macroblock type of a current macroblock is a prediction macroblock in view, and a value of 'mv_refinement_mode' is '2'.

Accordingly, the macroblocks of neighboring pictures become macroblocks MB_a and MB_b of neighboring pictures in the view direction by the macroblock type, and the macroblocks of pictures later in the view direction by the value of 'mv_refinement_mode'. The sum of the motion vector of and the 1/4 pixel refinement of the current macroblock is used as the motion vector of the current macroblock. The reference index is the same as the motion skip mode.

In other words, in FIG. 4A, 'mv_l0 (x, y) = mv_b_l0 (x + h, y + v) + mvd_refinement_l0 (x, y)' and 'mv_l1 (x, y) = mv_b_l1 (x + h, y + v) + mvd_refinement_l1 (x, y) ', and the reference index of MB_b becomes the reference index of the current macroblock (current MB). In FIG. 4 (b), 'mv_t_0 = mv_tb_0 + mvd_refinement_l0' and 'mv_t_1 = mv_tb_1 + mvd_refinement_l1', and the reference index of MB_b becomes the reference index of the current macroblock.

Finally, the motion vector predictor (hereinafter, referred to as MVP) prediction mode will be described. In the present invention, MVP prediction mode is a sum of the motion or disparity vector of the macroblock in the neighboring picture and the motion vector difference of the current macroblock as the motion or disparity vector of the current macroblock. Decoding a current macroblock by using a reference index of a macroblock in a neighboring picture as a reference index of the current macroblock. That is, to use the motion or disparity vector of the neighboring picture as a predictor of the motion or disparity vector for the current macroblock. To this end, in the present invention, 'mvp_prediction_model0' and 'mvp_prediction_model1' are newly defined.

For example, when the value of 'mvp_prediction_model0 (l1)' is '0', the macroblock type of the current macroblock is not used. That is, the MVP prediction mode according to the present invention is not used. When the value of 'mvp_prediction_model0 (l1)' is '1', it means forward prediction, and the macroblock of the picture that is forward in the time direction or the view direction of the picture to which the current macroblock belongs is The sum of the motion or disparity vector of MB_a, MB_c) and the motion vector difference of the current macroblock is used as the motion or disparity vector of the current macroblock. Here, in the present invention, 'mvd_l0' and 'mvd_l1' syntax in H.264 / AVC are used for the motion vector reference of the current macroblock. The reference index is the same as the motion skip mode. If the value of 'mvp_prediction_model0 (l1)' is '2', it means backward prediction, and the macroblock of the picture backward in the time direction or the view direction of the picture to which the current macroblock belongs. The sum of the motion or disparity vector of (MB_b, MB_d) and the motion or disparity vector of the current macroblock (MB_a, MB_c) and the motion vector difference of the current macroblock is used as the motion or disparity vector of the current macroblock. . The reference index is the same as the motion skip mode.

This will be described with reference to FIGS. 4 (a) and 4 (b). First, it is assumed that the macroblock type of the current macroblock is an inter-view prediction macroblock, and that the values of 'mvp_prediction_model0' and 'mvp_prediction_model1' are '2'. lets do it.

Therefore, the macroblocks of neighboring pictures become macroblocks MB_c and MB_d of neighboring pictures in the time direction by the macroblock type, and the macros of pictures later in the time direction by the values of 'mvp_prediction_model0' and 'mvp_prediction_model1'. The sum of the disparity vector of the block MB_d and the motion vector difference of the current macroblock is used as the disparity vector of the current macroblock. The reference index is the same as the motion skip mode.

In other words, in FIG. 4A, 'dis_l0 (x, y) = dis_b_l0 (x + h, y + v) + mvd_l0 (x, y)' and 'dis_l1 (x, y) = dis_b_l1 (x + h, y + v) + mvd_l1 (x, y) ', and the reference index of MB_d becomes the reference index of the current macroblock (current MB). In FIG. 4 (b), 'dis_v_0 = dis_vb_0 + mvd_l0' and 'dis_v_1 = dis_vb_1 + mvd_l1', and the reference index of MB_d becomes the reference index of the current macroblock.

5 and 6 are diagrams for describing the video signal decoding method of FIG. 3 (b) in detail and relate to motion prediction of the prediction flag information in FIG. 3 (b). 5 illustrates a case where the prediction method is a forward prediction method, and FIG. 6 illustrates a case where the prediction method is a backward prediction method.

First, the motion skip flag among the prediction flag information will be described. In the present invention, the motion skip flag is similar to the motion skip mode described above, but when only the forward prediction method is used, the motion skip flag uses the motion information of the macroblock of the picture in the time direction or the view direction, and only the backward prediction method is used. When used, it refers to decoding the current macroblock using motion information of a macroblock of a picture later in a time direction or a view direction. To this end, in the present invention, a syntax 'motion_skip_flag' is newly defined.

For example, when the value of 'motion_skip_flag' is '0', the macroblock type of the current macroblock is not used. That is, the motion skip flag according to the present invention is not used. In the present invention, when only the forward prediction method is used, the value of 'motion_ skip_flag' is '1' means forward prediction, and the forward motion is forwarded in the time direction or the view direction of the picture to which the current macroblock belongs. The motion or disparity vector of the macroblocks MB_a and MB_c of the picture in the forward) and the reference index are used as motion information of the current macroblock. In the present invention, when only the backward prediction method is used, the value of 'motion_skip_flag' is '1', which means backward prediction, and is backward in the time direction or view direction of the picture to which the current macroblock belongs. The motion or disparity vector of the macroblocks MB_b and MB_d of the picture in the backward direction and the reference index are used as motion information of the current macroblock.

This will be described with reference to FIGS. 5 (a) and 5 (b). First, it is assumed that the macroblock type of the current macroblock is a prediction macroblock in view, and the value of 'motion_skip_flag' is '1'.

Accordingly, the macroblocks of neighboring pictures become macroblocks MB_a and MB_b of neighboring pictures in the view direction by the macroblock type, and the macroblocks of the pictures previously in the view direction by the value of 'motion_skip_flag'. The motion vector and the reference index of are used as the motion vector and the reference index of the current macroblock.

In other words, in FIG. 5 (a), 'mv_l0 (x, y) = mv_a_l0 (x + h, y + v)' and 'mv_l1 (x, y) = mv_a_l1 (x + h, y + v)' The reference index of MB_a becomes the reference index of the current macroblock. In FIG. 5 (b), 'mv_t_0 = mv_ta_0' and 'mv_t_1 = mv_ta_1', and the reference index of MB_a becomes the reference index of the current macroblock.

6 (a) and 6 (b), it is assumed that a macroblock type of a current macroblock is an inter-view prediction macroblock, and a value of 'motion_skip_flag' is '1'.

Therefore, the macroblocks of neighboring pictures become macroblocks MB_c and MB_d of neighboring pictures in the time direction by the macroblock type, and the macroblocks of the pictures later in the time direction by the value of 'motion_skip_flag'. The motion vector and the reference index of are used as the motion vector and the reference index of the current macroblock.

In other words, in FIG. 6 (a), 'dis_l0 (x, y) = dis_b_l0 (x + h, y + v)' and 'dis_l1 (x, y) = dis_b_l1 (x + h, y + v)' The reference index of MB_d becomes the reference index of the current macroblock (current MB). In FIG. 6 (b), 'dis_v_0 = dis_vb_0' and 'dis_v_1 = dis_vb_1', and the reference index of MB_d becomes the reference index of the current macroblock.

Next, the motion refinement flag corresponds to the motion refinement mode and may be classified and used as the case where only the forward prediction method and the backward prediction method are used in the present invention, as described above.

Lastly, like the motion refinement flag, the MVP predicate flag corresponds to the MVP prediction mode, and only the forward prediction method and the backward prediction method are used in the present invention as the motion skip flag described above. Can be classified and used.

FIG. 7 is a diagram for describing the video signal decoding method of FIG. 3 (a) in detail and relates to texture prediction among prediction method information in FIG. 3 (a).

First, the residual prediction mode among the prediction method information will be described. In the present invention, the residual prediction mode refers to decoding a current macroblock by using a residual signal of a macroblock in a neighboring picture as a residual prediction signal of the current macroblock. To this end, the present invention newly defines a syntax called 'res_pred_mode'.

For example, when the value of 'res_pred_mode' is '0', the current macroblock is not decoded using the residual signal. That is, the residual prediction mode according to the present invention is not used. When the value of 'res_pred_mode' is '1', it means forward prediction, and the macroblocks MB_a and MB_c of the picture forward in the time direction or the view direction of the picture to which the current macroblock belongs. ) Is used as the residual prediction signal of the current macroblock. When the value of 'res_pred_mode' is '2', it means backward prediction, and the macroblock MB_b of the picture backward in the time direction or the view direction of the picture to which the current macroblock belongs , And the residual signal of MB_d) is used as the residual prediction signal of the current macroblock.

This will be described with reference to FIGS. 7A and 7B. First, suppose that the macroblock type of the current macroblock is an intra prediction macroblock, and the value of 'res_pred_mode' is '1'.

Accordingly, the macroblocks of neighboring pictures become macroblocks MB_c and MB_d of neighboring pictures in the temporal direction by the macroblock type, and the macroblocks of the pictures previously in the temporal direction by the value of 'res_pred_mode'. The residual signal of is used as the residual prediction signal of the current macroblock.

In other words, in FIG. 7 (a), 'res (x, y) = res' (x, y) + res_pred (x, y) = res' (x, y) + res (x + h, y + v) Becomes. Here, res (x, y) means the residual signal of the current macroblock, and res_pred (x, y) means the residual prediction signal of the current macroblock. Therefore, the residual prediction signal res_pred (x, y) of the current macroblock becomes 'res (x + h, y + v)', which is the residual signal of the macroblock MB_c of the picture preceding in the time direction. . Also, res' (x, y) is obtained from the encoder as the difference between the residual signal of the current macroblock and the residual prediction signal of the current macroblock.

In FIG. 7B, 'res = res_pred = res_MB_c'. Here, res means a residual signal of the current macroblock, and res_pred means a residual prediction signal of the current macroblock. In addition, res_MB_c means a residual signal of MB_c. That is, in FIG. 7B, the residual signal of the current macroblock is obtained from the residual signal of the neighboring macroblock.

Next, the intra prediction mode will be described. In the present invention, the intra prediction mode refers to decoding a current macroblock by using an intra signal of a macroblock in a neighboring picture as an intra signal of a current macroblock. To this end, the present invention newly defines a syntax of 'intra_pred_mode'.

For example, when the value of 'intra_pred_mode' is '0', the current macroblock is not decoded using the intra signal. That is, the intra prediction mode according to the present invention is not used. When the value of 'intra_pred_mode' is '1', it means forward prediction, and the macroblocks MB_a and MB_c of the picture forward in the time direction or the view direction of the picture to which the current macroblock belongs. ) Is used as the intra signal of the current macroblock. When the value of 'intra_pred_mode' is '2', it means backward prediction, and the macroblock of the picture that is backward in the time direction or the view direction of the picture to which the current macroblock belongs is MB_b, The intra signal of MB_d) is used as the intra signal of the current macroblock.

This will be described with reference to FIGS. 7A and 7B. First, it is assumed that a macroblock type of a current macroblock is an intra prediction macroblock and a value of 'intra_pred_mode' is '1'.

In this case, the macroblocks of neighboring pictures become macroblocks MB_c and MB_d of neighboring pictures in the time direction by the macroblock type, and the macroblocks (MB_c) of pictures that exist before in the time direction by the value of 'intra_pred_mode'. ) Is used as the intra signal of the current macroblock.

Also, assume that the macroblock type of the current macroblock is an inter-view prediction macroblock, and the value of 'intra_pred_mode' is '2'.

In this case, the macroblocks of neighboring pictures by the macroblock type become the macroblocks MB_a and MB_b of neighboring pictures in the view direction, and the macroblocks of the pictures later in the view direction by the value of 'intra_pred_mode'. ) Is used as the intra signal of the current macroblock.

8 and 9 are diagrams for describing the video signal decoding method of FIG. 3B in detail, and refer to texture prediction of the prediction flag information in FIG. 3B. 8 illustrates a case where the prediction method is a forward prediction method, and FIG. 9 illustrates a case where the prediction method is a backward prediction method.

First, the residual prediction flag among the prediction flag information will be described. In the present invention, the residual prediction flag is similar to the residual prediction mode described above, but when only the forward prediction method is used, the residual signal of the macroblock of the picture in the time direction or the view direction is used. When only the word prediction method is used, it means decoding the current macroblock by using the residual signal of the macroblock of the picture later in the time direction or the view direction. To this end, the present invention newly defines a syntax called 'res_pred_flag'.

For example, when the value of 'res_pred_flag' is '0', the current macroblock is not decoded using the residual signal. That is, the residual prediction flag according to the present invention is not used. In the present invention, when only the forward prediction method is used, the value of 'res_pred_flag' means '1', which means forward prediction, and forward in the time direction or the view direction of the picture to which the current macroblock belongs. The residual signals of the macroblocks MB_a and MB_c of the picture in &quot;) are used as the residual prediction signals of the current macroblock. In the present invention, when only the backward prediction method is used, the value of 'res_pred_flag' means '1', which means backward prediction, and is later in the time direction or the view direction of the picture to which the current macroblock belongs. The residual signal of the macroblocks MB_b and MB_d of the picture in (backward) is used as the residual prediction signal of the current macroblock.

This will be described with reference to FIGS. 8 (a) and 8 (b). First, it is assumed that a macroblock type of a current macroblock is a prediction macroblock in view and a value of 'res_pred_flag' is '1'.

Therefore, the macroblocks of neighboring pictures become macroblocks MB_c and MB_d of neighboring pictures in the time direction according to the macroblock type, and the macroblocks of the pictures preceding in the time direction by the value of 'res_pred_flag'. The residual signal of is used as the residual prediction signal of the current macroblock.

In other words, in FIG. 8 (a), 'res (x, y) = res' (x, y) + res_pred (x, y) = res' (x, y) + res (x + h, y + v) Becomes. Here, res (x, y) means the residual signal of the current macroblock, and res_pred (x, y) means the residual prediction signal of the current macroblock. Therefore, the residual prediction signal res_pred (x, y) of the current macroblock becomes 'res (x + h, y + v)', which is the residual signal of the macroblock MB_c of the picture preceding in the time direction. . Also, res' (x, y) is obtained from the encoder as the difference between the residual signal of the current macroblock and the residual prediction signal of the current macroblock.

In FIG. 8 (b), 'res = res_pred = res_MB_c'. Here, res means a residual signal of the current macroblock, and res_pred means a residual prediction signal of the current macroblock. In addition, res_MB_c means a residual signal of MB_c. That is, in FIG. 8B, the residual signal of the current macroblock is obtained from the residual signal of the neighboring macroblock.

9 (a) and 9 (b), it is assumed that a macroblock type of a current macroblock is an inter-view prediction macroblock, and a value of 'res_pred_flag' is '1'.

Therefore, the macroblocks of neighboring pictures become macroblocks MB_a and MB_b of neighboring pictures in the view direction by the macroblock type, and the macroblocks of the pictures later in the view direction by the value of 'res_pred_flag'. The residual signal of is used as the residual prediction signal of the current macroblock.

In other words, in FIG. 9 (a), 'res (x, y) = res' (x, y) + res_pred (x, y) = res' (x, y) + res (x + h, y + v) Becomes. Here, res (x, y) means the residual signal of the current macroblock, and res_pred (x, y) means the residual prediction signal of the current macroblock. Therefore, the residual prediction signal res_pred (x, y) of the current macroblock becomes 'res (x + h, y + v)' which is the residual signal of the macroblock MB_b of the picture later in the view direction. . Also, res' (x, y) is obtained from the encoder as the difference between the residual signal of the current macroblock and the residual prediction signal of the current macroblock.

In FIG. 9 (b), 'res = res_pred = res_MB_d'. Here, res means a residual signal of the current macroblock, and res_pred means a residual prediction signal of the current macroblock. In addition, res_MB_d means a residual signal of MB_d. That is, in FIG. 8B, the residual signal of the current macroblock is obtained from the residual signal of the neighboring macroblock MB_d.

Next, the intra prediction flag corresponds to the intra prediction mode, and may be classified and used as the case where only the forward prediction method and the backward prediction method are used in the present invention, such as the residual prediction flag described above. have.

10 is a flowchart illustrating another embodiment of a video signal decoding method according to the present invention.

As shown in FIG. 10, first, the prediction mode is read (S101). Here, in the present invention, the prediction mode is whether the current macroblock is a general macroblock without using the video signal decoding method according to the present invention or a macroblock by intra-view prediction using the video signal decoding method according to the present invention. Indicates whether it is a macroblock by recognition or inter-view prediction.

Next, the prediction method information of the current macroblock is read (S103). Here, the prediction method information according to the present invention may be divided into three pieces of information on motion prediction and two pieces of information on texture prediction.

Finally, the current macroblock is decoded according to the read prediction mode and the prediction scheme information (S105). In the video signal decoding method according to the present invention, when the current macroblock is a macroblock that is not based on intra-view prediction and inter-view prediction by the read prediction mode, the current macroblock type is read and the current macroblock is decoded. When the current macroblock is a macroblock by intra-view prediction or inter-view prediction, the current macroblock is decoded using macroblock information of a picture neighboring in the time direction or the view direction. Here, macroblock information of a neighboring picture may be divided into motion information and texture information. Also, according to the read prediction information, macroblocks of pictures neighboring in the time direction or the view direction are macroblocks of pictures forward in the time direction or the view direction, or backward in the time direction or the view direction. This is the macroblock of the picture in. In other words, the prediction mode may determine whether the current macroblock and the neighboring picture are the neighboring pictures in the time direction or the neighboring pictures in the view direction, and the prediction method information may be used in all directions in the time direction or the view direction. It can be seen whether the picture is in the forward or backward.

11 is a block diagram showing an embodiment of a video signal decoding apparatus according to the present invention. The video signal decoding apparatus according to the present invention corresponds to the decoding unit 50 in FIG. 2 and includes a macroblock type reading unit 111, a prediction method information reading unit 113, and a decoding unit 115.

The macroblock type reader 111 reads the macroblock type of the current macroblock from the received bitstream. In the present invention, the macroblock type indicates whether the current macroblock is a macroblock by intra-view prediction or a macroblock by inter-view prediction.

The prediction method information reading unit 113 reads prediction method information of the current macroblock. In the present invention, the prediction method information may be divided into information about motion prediction and texture prediction. As described above, information about motion prediction includes a motion skip mode, a motion refinement mode, and an MVP prediction mode, and information about texture prediction includes a residual prediction mode and an intra prediction mode.

The decoder 115 decodes the current macroblock using the macroblock type and the prediction method information. In the present invention, the decoder 115 decodes the current macroblock using macroblock information of neighboring pictures in the time direction or the view direction according to whether the macroblock type is an intra-view prediction macroblock or an inter-view prediction macroblock. . Here, the macroblock information of a neighboring picture is a reference index, a motion or a disparity vector of a macroblock of a neighboring picture in case of motion prediction, and a residual signal of a macroblock of a neighboring picture in the case of texture prediction, It can be an intra signal. In addition, the macroblocks of pictures neighboring in the time direction or the view direction may be macroblocks of pictures that are before or after the time direction or the view direction according to prediction method information.

12 (a) and 12 (b) are flowcharts illustrating still another embodiment of a video signal decoding method according to the present invention.

As shown in FIG. 12 (a), first, a pixel-based disparity vector is obtained from a disparity map (S121). The disparity map in the present invention is also called a depth map, and means a map indicating the positional difference of pixels between two viewpoints in the multi-view video coding of the present invention. On the other hand, in the in-view prediction of multi-view video coding, the disparity vector should be transmitted to the decoder for disparity compensation of the current macroblock, but in the present invention, when the disparity map is transmitted to the decoder, The disparity vector is obtained using the parity map.

Next, a block unit disparity vector is derived from a pixel unit disparity vector (S122). In this case, the pixel-based disparity vector may be derived as a 4 × 4 block-based disparity vector by a method such as an average value or a median.

Finally, the current macroblock is decoded using the disparity vector that is grouped by combining blocks having the same disparity vector among the derived block unit disparity vectors (S124). Herein, the 4 × 4 block unit disparity vector derived in step 122 is combined into 8 × 8 block unit disparity vectors by grouping blocks having the same disparity vector. Subsequently, the 8 × 8 block unit disparity vector is combined into a 16 × 16 or 16 × 8 or 8 × 16 block unit disparity vector by grouping an 8 × 8 block unit disparity vector having the same disparity vector. As a result, the current macroblock is decoded using disparity compensation by the combined block-by-block disparity vector.

12 (b) is a flowchart of an embodiment of a method of obtaining the pixel-based disparity vector of step 120. First, a residual disparity vector is obtained from the residual disparity map (S121). Here, the residual disparity map is obtained from the encoding stage, which will be described in detail with reference to FIG. 13.

Next, the pixel-based disparity vector is obtained by adding the residual disparity vector and the block-based disparity vector. Here, the block-by-block disparity vector is obtained by block matching motion estimation.

FIG. 13 is a diagram illustrating an embodiment of a video signal decoding method of FIG. 12.

FIG. 13A illustrates a relationship between the disparity vector and the disparity map, and it can be seen that the disparity vector can be obtained as the disparity map. As shown, dis_l0 can be obtained from the disparity map A and dis_l1 can be obtained from the disparity map B.

FIG. 13B is a diagram illustrating an embodiment of the video signal decoding method of FIG. 12A. As shown, STEP 1 obtains pel disparity vectors from the disparity map.

In STEP 2, block disparity vectors are derived from the pixel-by-pixel disparity vector. Here, the block unit disparity vector becomes a 4 × 4 block unit disparity vector.

In STEP 3, the 4 × 4 block disparity vector derived in STEP 2 is combined into 8 × 8 block unit disparity vectors by grouping blocks having the same disparity vector. Here, the reason for grouping blocks having the same disparity vector is a vector representing a positional difference of pixels between viewpoints, that is, a disparity vector has the same or similar disparity vector in the case of adjacent blocks. This will increase.

Subsequently, the 8 × 8 block unit disparity vector is combined into a 16 × 16 or 16 × 8 or 8 × 16 block unit disparity vector by grouping an 8 × 8 block unit disparity vector having the same disparity vector. As a result, the current macroblock is decoded using disparity compensation by the combined block-by-block disparity vector.

FIG. 13 (c) is a diagram showing an embodiment of the video signal decoding method of FIG. 12 (b). In the encoder, a pixel-by-pixel disparity vector (pel dis. Vec.) Is obtained from a disparity map. Next, a residual disparity vector (residu. Dis. Vec.) That is a difference between the pixel-based disparity vector and the block disparity vector (block dis. Vec.) By block matching motion estimation is obtained. . Finally, a residual disparity map is obtained from the residual disparity vector.

The decoder obtains the pixel-by-pixel disparity vector from the residual disparity map in the reverse order of the method described in the encoder.

In summary, in the video signal decoding method according to the present invention of FIG. 12A, when a disparity map is transmitted from an encoder to a decoder, a disparity vector is obtained using the disparity map. On the other hand, the video signal decoding method of FIG. 12 (b) transmits the residual disparity map to the decoding stage instead of the disparity map and obtains the pixel-by-pixel disparity vector, thereby reducing the transmission amount and improving coding efficiency.

As mentioned above, preferred embodiments of the present invention are disclosed for purposes of illustration, and those skilled in the art can improve and change various other embodiments within the spirit and technical scope of the present invention disclosed in the appended claims below. , Replacement or addition would be possible.

As described above, in the video signal decoding method and apparatus according to the present invention, various prediction methods capable of removing redundant information between viewpoints in multi-view video coding are possible. By using the disparity map to remove, the coding efficiency is increased.

Claims (25)

In the video signal decoding method using inter-view prediction, Reading a macroblock type of a current macroblock; And And decoding the current macroblock using macroblock information of a neighboring picture according to the read macroblock type. The method of claim 1, wherein the video signal decoding method is Reading the prediction method information of the current macroblock; And the macroblock of the neighboring picture is a macroblock of a picture which is before or after in a time direction or a view direction according to the read prediction information. The method of claim 2, wherein the macroblock of the neighboring picture is And a macroblock of a position corresponding to the current macroblock. 4. The method of claim 3, wherein the macroblock at a position corresponding to the current macroblock is found using a motion or disparity vector of the current macroblock. The method of claim 2, wherein the macroblock of the neighboring picture is And a macroblock at the same position as the current macroblock. The method of claim 3 or 5, wherein the video signal decoding method Reading a reference index and a motion or disparity vector of macroblocks of neighboring pictures in the time direction or the view direction according to whether the read macroblock type is an intra-view prediction macroblock or an inter-view prediction macroblock; The video signal decoding method further comprising. The picture of claim 6, wherein the neighboring pictures in the time direction or the view direction are arranged. The read macroblock type is a picture neighboring in the view direction when the macroblock is in the viewpoint, and the read macroblock type is a picture neighboring in the temporal direction in the case of the inter-view prediction macroblock. A video signal decoding method. 8. The method of claim 7, wherein decoding the current macroblock comprises: Generating a reference index of the current macroblock using the read reference index; And Generating a motion or disparity vector of the current macroblock using the read motion or disparity vector. 8. The method of claim 7, wherein decoding the current macroblock comprises: Generating a reference index of the current macroblock using the read reference index; And Generating a motion or disparity vector of the current macroblock using the sum of the read motion or disparity vector and the quarter-pixel refinement of the current macroblock. . 8. The method of claim 7, wherein decoding the current macroblock comprises: Generating a reference index of the current macroblock using the read reference index; And Generating a motion or disparity vector of the current macroblock using the sum of the read motion or disparity vector and the motion vector difference of the current macroblock. The method of claim 3 or 5, wherein the video signal decoding method And reading out a residual signal of a macroblock of a neighboring picture in the time direction or the view direction according to whether the read macroblock type is an intra-view prediction macroblock or an inter-view prediction macroblock. Video signal decoding method. 12. The picture of claim 11, wherein the neighboring picture in the time direction or the view direction is The read macroblock type is a picture neighboring in the temporal direction when the macroblock is in the view, and the read macroblock type is a picture neighboring in the view direction in the case of the inter-view prediction macroblock. A video signal decoding method. 13. The method of claim 12, wherein decoding the current macroblock comprises: And generating a residual signal of the current macroblock by using the read residual signal. The method of claim 3 or 5, wherein the video signal decoding method And reading an intra signal of a macroblock of a neighboring picture in the time direction or the view direction according to whether the read macroblock type is an intra-view prediction macroblock or an inter-view prediction macroblock. Video signal decoding method. The picture of claim 14, wherein the neighboring picture in the time direction or the view direction is selected. The read macroblock type is a picture neighboring in the temporal direction when the macroblock is in the view, and the read macroblock type is a picture neighboring in the view direction in the case of the inter-view prediction macroblock. A video signal decoding method. 16. The method of claim 15, wherein decoding the current macroblock comprises: And generating an intra signal of the current macroblock by using the read intra signal. The method of claim 1, wherein the video signal decoding method is Reading prediction flag information indicating whether to apply the prediction method; The macroblock of the neighboring picture is a macroblock of a picture preceding in a time direction or a view direction when the prediction method is a forward prediction method, and when the prediction method is a backward prediction method. And a macroblock of a picture later in the time direction or the view direction. In the video signal decoding method using inter-view prediction, Reading a prediction mode indicating whether the current macroblock is a macroblock by intra-view or inter-view prediction; And And decoding the current macroblock using macroblock information of a neighboring picture in accordance with the read prediction mode. 19. The method of claim 18, wherein the video signal decoding method is Reading the prediction method information of the current macroblock; And the macroblock of the neighboring picture is a macroblock of a picture before or after the time direction or the view direction according to the read prediction information. In the video signal decoding method using inter-view prediction, Obtaining a pixel-by-pixel disparity vector; Deriving a block unit disparity vector from the obtained pixel unit disparity vector; And And decoding a current macroblock using a disparity vector that is combined by grouping blocks having the same disparity vector among the derived block unit disparity vectors. The pixel-based disparity vector of claim 18, wherein The video signal decoding method characterized by obtaining from a disparity map. 19. The method of claim 18, wherein obtaining the pixel-based disparity vector is Obtaining the residual disparity vector from the residual disparity map; And And calculating the pixel-based disparity vector by adding the obtained residual disparity vector and the block-based disparity vector based on block matching motion estimation. In the video signal decoding apparatus using inter-view prediction, A macroblock type reader for reading a macroblock type of a current macroblock; A prediction method information reading unit which reads prediction method information of the current macroblock; And And a decoder which decodes the current macroblock by using the macroblock type and the prediction scheme information. The method of claim 23, wherein the decoding unit And decoding the current macroblock using macroblock information of neighboring pictures in a time direction or a view direction according to whether the macroblock type is an intra-view prediction macroblock or an inter-view prediction macroblock. . The method of claim 24, And a macroblock of a picture neighboring in the time direction or the view direction is a macroblock of a picture before or after the time direction or the view direction according to the read prediction information.

Images