CN114170331A - A kind of image data compression method and system based on artificial intelligence - Google Patents
A kind of image data compression method and system based on artificial intelligence Download PDFInfo
- Publication number
- CN114170331A CN114170331A CN202111471673.1A CN202111471673A CN114170331A CN 114170331 A CN114170331 A CN 114170331A CN 202111471673 A CN202111471673 A CN 202111471673A CN 114170331 A CN114170331 A CN 114170331A
- Authority
- CN
- China
- Prior art keywords
- color information
- dimensional
- color
- compressed
- difference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
- G06T9/002—Image coding using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention relates to the technical field of artificial intelligence, in particular to an image data compression method and system based on artificial intelligence. The method comprises the steps of analyzing two-dimensional color information of a training image, separating boundary two-dimensional color information reaching a lossless effect boundary according to two-dimensional standard difference of the two-dimensional color information on a preset order, obtaining a Gaussian model of each pixel point according to the two-dimensional standard difference, obtaining color tropism according to the difference of compressed color information and the boundary two-dimensional color information on the Gaussian model, and further combining the color difference of other pixel points in a neighborhood range before and after pixel point compression to construct a loss function. And training the self-coding neural network by using the two-dimensional color information as training data. The lossless compression is realized through the self-coding neural network. The invention realizes lossless compression of image data by using the self-coding neural network by restricting the network by obtaining the boundary of color tolerance.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an image data compression method and system based on artificial intelligence.
Background
A common data compression method is to perform a complicated inverse operation on an image to achieve lossless compression. The inverse operation is complicated and the operation difficulty is large for different types of image data. With the continuous development of the field of artificial intelligence, in the prior art, a network can be constructed through artificial intelligence, image data can be compressed fast in a self-adaptive mode, and a common network comprises a self-coding neural network and the like.
In the process of compressing image data by using a network, lossless compression of the image is expected to be realized, namely color information in the image is unchanged before and after compression, for the image data, the realization of complete lossless compression is complex, the difficulty is high, the network construction cost is increased, and the compression efficiency is influenced.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide an image data compression method and system based on artificial intelligence, wherein the adopted technical scheme is as follows:
the invention provides an image data compression method based on artificial intelligence, which comprises the following steps:
acquiring a plurality of training images; converting the training image into a CIE-xyY color space to obtain two-dimensional color information;
taking the two-dimensional color information as training data of a self-coding neural network; the self-coding neural network outputs compressed color information; performing Gaussian fitting according to the two-dimensional standard deviation of the two-dimensional color information of each pixel point in the training image on a preset order to obtain a Gaussian model of each pixel point; boundary two-dimensional color information of the two-dimensional color information is obtained according to the two-dimensional standard deviation; obtaining a color tendency according to the difference between the compressed color information and the boundary two-dimensional color information on the Gaussian model; taking the difference between the two-dimensional color information of the pixel point and the two-dimensional color information of other pixel points in a preset neighborhood range as a first color difference; taking the difference of the compressed color information of the pixel point and other pixel points in the neighborhood range as a second color difference; constructing a loss function of the self-coding neural network according to the difference between the first color difference and the second color difference and the color tendency; training the self-coding neural network according to the loss function and the training data;
and inputting the two-dimensional color information of the image to be compressed into the self-coding neural network, and outputting image compression data.
Further, the preset order is set to be three orders.
Further, the performing gaussian fitting according to the two-dimensional standard deviation of the two-dimensional color information of each pixel point in the training image on a preset order includes:
and carrying out Gaussian fitting by taking the two-dimensional color information of the pixel points as a center and taking the two-dimensional standard deviation as a model mean value to obtain the Gaussian model.
Further, the obtaining boundary two-dimensional color information of the two-dimensional color information according to the two-dimensional standard deviation comprises:
the boundary two-dimensional color information comprises negative boundary two-dimensional color information and positive boundary two-dimensional color information; subtracting the two-dimensional standard deviation from the two-dimensional color information to obtain negative boundary two-dimensional color information; and adding the two-dimensional color information and the standard deviation to obtain positive boundary two-dimensional color information.
Further, the obtaining a color tendency according to a difference between the compressed color information and the boundary two-dimensional color information on the gaussian model comprises:
obtaining the color tropism according to a color tropism formula; the color tendency formula includes:
wherein, qxiThe color tropism, G, of the ith pixel pointi(yi') is a model value of said compressed color information of said i-th pixel point on said Gaussian model, (X)i，Yi) The two-dimensional color information of the ith pixel point,
the two-dimensional standard deviation of the ith pixel point,
and the model value of the boundary two-dimensional color information of the ith pixel point on the Gaussian model is obtained.
Further, the constructing the loss function of the self-coding neural network according to the difference between the first color difference and the second color difference and the color tendency comprises:
the loss function includes:
wherein loss is the loss function, I is the total number of the pixel points, qxiThe color tropism of the ith pixel point, S is the neighborhood range, yiThe two-dimensional color information, m, for the ith pixel pointjIs the jth other in the neighborhoodThe two-dimensional color information of the pixel points, E (y)i,mj) Is the first color difference, y'iM 'of the compressed color information of the ith pixel point'jThe compressed color information, E (y'i,m′j) Is the second color difference.
Further, the inputting the two-dimensional color information of the image to be compressed into the self-coding neural network, and outputting the image compression data further comprises
Acquiring gray information of the image to be compressed, inputting the gray information of the image to be compressed and the two-dimensional color information into the self-coding neural network, and outputting composite image compressed data; the composite image compressed data includes the image compressed data and the gradation information.
Further, the method further includes, after outputting the composite image compressed data:
decoding the composite image compressed data to obtain color information and the gray information; and replacing the gray information with the brightness component information in the color information to obtain a restored image.
The invention also provides an artificial intelligence based image data compression system, which comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, and is characterized in that the processor realizes any step of the artificial intelligence based image data compression method when executing the computer program.
The invention has the following beneficial effects:
the embodiment of the invention analyzes the two-dimensional color information of the image and obtains the boundary two-dimensional color information according to the two-dimensional standard deviation of the two-dimensional color information. The purpose of obtaining the two-dimensional standard deviation is to obtain a macadam ellipse of the two-dimensional color information, and the boundary two-dimensional color information is the boundary of the macadam ellipse. And further obtaining color tropism, wherein the color tropism is used as a part of a loss function, so that the compressed color information can be in the boundary color information, namely in the MacAdam ellipse, the color difference value which can be perceived by human eyes cannot be caused, and lossless compression is realized. Further, the contrast between one pixel point and the surrounding pixel points in the compressed color information can be close to that before compression by using the difference between the first color difference and the second color difference as a part of the loss function through the color difference relationship between the first color difference and the second color difference resolution pixel points and the surrounding pixel points, and the quality of the compressed color information is ensured.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of an artificial intelligence-based image data compression method according to an embodiment of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description of the embodiments, structures, features and effects of the image data compression method and system based on artificial intelligence according to the present invention will be made with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following describes a specific scheme of an image data compression method and system based on artificial intelligence in detail with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of an artificial intelligence based image data compression method according to an embodiment of the present invention is shown, where the method includes:
step S1: acquiring a plurality of training images; and converting the training image into a CIExyY color space to obtain two-dimensional color information.
A common color metric model of image data is typically an RGB model, i.e., colors of an image are represented according to channel values in R, G, B three channels. However, the RGB color model is a color space related to devices, and each device has different definitions when using the RGB model, so that the color effect is different, and thus the definitions of RGB between devices are not universal. Therefore, in image data compression, directly compressing RGB color information results in more obvious differences in color effects of the compressed color information in different devices. Therefore, a plurality of training images are converted into a CIE-xyY color space, the CIE-xyY color space is an imaginary primary color system related to RGB, a standard light source is defined, and the method is more suitable for color calculation. It should be noted that the CIE-xyz color space is converted from CIE-xyz, that is, the RGB color model is first converted into the CIE-xyz color space, and then converted from the CIE-xyz color space to the CIE-xyz color space, where the conversion relationship is a well-known technology and is not described herein. Finally obtaining two-dimensional color information y of training image in CIE-xyY color spacei＝[Xi,Yi]And i represents the ith pixel point in the training image.
Step S2: taking two-dimensional color information as training data of a self-coding neural network; outputting the self-coding neural network as compressed color information; performing Gaussian fitting according to the two-dimensional standard deviation of the two-dimensional color information of each pixel point in the training image on a preset order to obtain a Gaussian model of each pixel point; boundary two-dimensional color information of the two-dimensional color information is obtained according to the two-dimensional standard deviation; obtaining color tropism according to the difference between the compressed color information and the boundary two-dimensional color information on a Gaussian model; taking the difference of two-dimensional color information of the pixel point and other pixel points in a preset neighborhood range as a first color difference; taking the difference of the compressed color information of the pixel point and other pixel points in the neighborhood range as a second color difference; constructing a loss function of the self-coding neural network according to the difference between the first color difference and the second color difference and the color tendency; the self-coding neural network is trained based on the loss function and the training data.
The two-dimensional color information of the training image is used as training data of the self-coding neural network, and because each pixel point comprises the two-dimensional color information, the input data volume of the self-coding neural network is twice of the image size. In the embodiment of the invention, the self-coding neural network adopts a full-connection network structure, expands the two-dimensional color information of the image to form one-dimensional input data and outputs the compressed color information yi', where i denotes the ith pixel point in the training image. It should be noted that the self-coding neural network for image data compression is a known technology in the art, and the details of the image data compression related method in the self-coding neural network are not repeated herein.
When the common self-coding neural network performs image compression, various complex processing means are used to retain more color information in order to ensure lossless compression of image color information. However, since the self-coding neural network is essentially a fitter, the color information of all pixel points before and after image compression cannot be consistent in percentage, and too many processing methods are used, so that the network iteration times are increased, resources are consumed in a transition manner, and the cost of a compression task is increased. It should be noted that, the change of the color information of the image has color tolerance, that is, the degree of change of the color information does not change in color respectively by human eyes within the color tolerance, so that when the self-coding neural network performs image compression, the fitting requirement can be relaxed, and the change of the color information before and after compression is within the color tolerance, that is, the color of the image is considered not to change, thereby implementing lossless compression of the image.
The MacAdam ellipse is a closed shape formed in the process that the analysis pixel point and the pixel point have color difference in different directions. The standard deviation of the color information of the pixel point represented by the MacAdam ellipse and other pixel points represents the resolution of chromaticity, namely the first-order MacAdam ellipse refers to the standard deviation of the difference of the color matching result which is one time of the color information of the target pixel point, and the first-order MacAdam ellipse refers to the standard deviation of the difference of the color matching result which is one time of the color information of the target pixel pointThe other orders are the same as the first order. The color information of the pixel points in the first-order MacAdam ellipse and the second-order MacAdam ellipse can not be distinguished by human eyes, and the pixel points in the first-order MacAdam ellipse and the second-order MacAdam ellipse can be considered to be the same color. The color corresponding to the boundary of the third-order MacAdam ellipse is different from the color corresponding to the center point, so that the third-order MacAdam ellipse can be used as a boundary, namely the preset order is set to be third order, and the two-dimensional color information [ X ] of each pixel point in the training imagei,Yi]Three-order two-dimensional standard deviation of
The analysis was performed as above.
Because the ellipse equation is complex, the amount of calculation is complex when the ellipse equation is used for analyzing whether the compressed color information is in the range of the third-order MacAdam ellipse. Since the distribution of the distances between the pixel points and the center of the ellipse in the macadam ellipse is approximately gaussian, the position of the compressed color information can be analyzed in the gaussian distribution model. The obtaining of the gaussian model of each pixel point by performing gaussian fitting according to the two-dimensional standard deviation specifically includes:
and performing Gaussian fitting by taking the two-dimensional color information of the pixel points as a center and taking the two-dimensional standard deviation as a model mean value to obtain a Gaussian model.
Because the compressed color information needs to be within the third-order macadam ellipse and not on the ellipse boundary, the boundary of the macadam ellipse needs to be obtained, and the boundary of the two-dimensional color information can be obtained according to the two-dimensional standard deviation, which specifically includes:
the boundary two-dimensional color information includes negative boundary two-dimensional color information and positive boundary two-dimensional color information. And subtracting the two-dimensional standard deviation from the two-dimensional color information to obtain the negative boundary two-dimensional color information. And adding the two-dimensional color information and the standard deviation to obtain the positive boundary two-dimensional color information. That is, the boundary two-dimensional color information of the ith pixel point is
The boundary two-dimensional color information is embodied in MacAdam ellipse as major axis and minor axis of ellipseIntersection points on the circle boundaries.
Because the central point of the macadam ellipse is two-dimensional color information, the closer the compressed color information is to the central point in the macadam ellipse, the better the color information, therefore, the compressed color information needs to be ensured to be close to the two-dimensional color information during fitting in the self-coding neural network, so that only the extreme case of the boundary two-dimensional color information needs to be analyzed, and other points on the boundary of the ellipse do not need to be considered.
When the difference between the compressed color information and the boundary two-dimensional color information on the Gaussian model is greater than zero, the current compressed color information can be considered to be in a third-order MacAdam ellipse, and the larger the difference is, the closer the difference is to the central point of the Gaussian model, so that the color tendency qx can be obtained according to the difference between the compressed color information and the boundary two-dimensional color information on the Gaussian modeliThe method specifically comprises the following steps:
and obtaining the color tropism according to a color tropism formula. The color tendency formula includes:
wherein, qxiIs the color tendency of the ith pixel point, Gi(y′i) Is the model value of the compressed color information of the ith pixel point on the Gaussian model, (X)i，Yi) Is the two-dimensional color information of the ith pixel point,
is the two-dimensional standard deviation of the ith pixel point,
and obtaining a model value of the boundary two-dimensional color information of the ith pixel point on the Gaussian model. In the color tendency formula, data are corrected through a-exp () function, when the model values are excessively different, the color tendency is a number which tends to 0, the phenomenon that the model values are excessively close to a central point, the network fitting times are increased, and network resources are wasted is avoided. And the compression is represented by the relationship of negative correlation that the smaller the color tendency isThe better the quality of the color information.
The network is constrained according to the color tendency, so that the effect presented by the compressed color information can meet the lossless state in human eyes, but the lossless state at the moment is only the state of the pixel points before and after compression, and for an image main body, if the color information between the pixel points has a large difference, the large difference can still cause the large difference in vision before and after image compression. Therefore, the color information difference between the pixel point and other pixel points in the preset neighborhood before and after compression needs to be analyzed, and the method specifically comprises the following steps:
and taking the difference of the two-dimensional color information of the pixel point and other pixel points in the preset neighborhood range as a first color difference. And taking the difference of the compressed color information of the pixel point and other pixel points in the neighborhood range as a second color difference. The first color difference and the second color difference before and after compression should be similar, so that the difference between the first color difference and the second color difference can be obtained to represent the similar degree, i.e. the smaller the difference is, the closer the two are, the better the compression effect is. In the embodiment of the present invention, the neighborhood range is set to 8 neighborhoods.
Further constructing a loss function from the encoded neural network in combination with the color tendency, wherein the loss function comprises:
wherein loss is a loss function, I is the total number of pixel points, qxiIs the color tropism of the ith pixel point, S is the neighborhood range, yiIs the two-dimensional color information of the ith pixel point, mjTwo-dimensional color information for the jth other pixel point in the neighborhood, E (y)i,mj) Is of a first color difference, y'iIs compressed color information of the ith pixel point, m'jCompressed color information for the jth other pixel point in the neighborhood range, E (y'i,m′j) Is the second color difference.
The loss value of the network in each iteration updating can be analyzed according to the loss function, and the smaller the loss value is, the better the compression quality is. And finishing the training of the network when the loss value is stable and minimum by a fitting method such as a gradient descent method.
Step S3: and inputting the two-dimensional color information of the image to be compressed into a coding neural network, and outputting image compression data.
After training of the training data and the loss function, the two-dimensional color information of the image to be compressed can be directly input into the self-coding neural network, and the image compression data is output. It should be noted that the loss function obtained in the embodiment of the present invention may be trained in combination with the loss function of other conventional image compression self-coding neural networks, so as to further achieve a better compression effect.
Preferably, the gray information of the image to be compressed is acquired, the gray information and the two-dimensional color information of the image to be compressed are input into the self-coding neural network, and the composite image compressed data is output. The composite image compressed data includes image compressed data and gray scale information the composite image compressed data is available for data storage and data transmission. When the composite image compressed data is required to be subjected to image reduction, decoding the composite image compressed data to obtain color information and gray information; and replacing the gray information with the brightness component information in the color information to obtain a restored image. The color information difference between the restored image and the image to be compressed cannot be observed by human eyes, namely the restored image is a lossless image in human eye vision, and lossless compression of the image is realized.
In summary, the embodiment of the present invention analyzes the two-dimensional color information of the training image, separates out boundary two-dimensional color information reaching the lossless effect limit according to the two-dimensional standard difference of the two-dimensional color information on the preset order, obtains a gaussian model of each pixel point according to the two-dimensional standard difference, obtains a color trend according to the difference between the compressed color information and the boundary two-dimensional color information on the gaussian model, and further constructs a loss function by combining the color differences between the compressed color information and other pixel points in the neighborhood range before and after the compression of the pixel points. And training the self-coding neural network by using the two-dimensional color information as training data. The lossless compression is realized through the self-coding neural network. The embodiment of the invention restricts the network by obtaining the boundary of the color tolerance, and realizes the lossless compression of the image data by using the self-coding neural network.
The invention also provides an artificial intelligence based image data compression system, which comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, and is characterized in that the processor realizes any step of the artificial intelligence based image data compression method when executing the computer program.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (9)
1. An artificial intelligence based image data compression method, the method comprising:
acquiring a plurality of training images; converting the training image into a CIE-xyY color space to obtain two-dimensional color information;
taking the two-dimensional color information as training data of a self-coding neural network; the self-coding neural network outputs compressed color information; performing Gaussian fitting according to the two-dimensional standard deviation of the two-dimensional color information of each pixel point in the training image on a preset order to obtain a Gaussian model of each pixel point; boundary two-dimensional color information of the two-dimensional color information is obtained according to the two-dimensional standard deviation; obtaining a color tendency according to the difference between the compressed color information and the boundary two-dimensional color information on the Gaussian model; taking the difference between the two-dimensional color information of the pixel point and the two-dimensional color information of other pixel points in a preset neighborhood range as a first color difference; taking the difference of the compressed color information of the pixel point and other pixel points in the neighborhood range as a second color difference; constructing a loss function of the self-coding neural network according to the difference between the first color difference and the second color difference and the color tendency; training the self-coding neural network according to the loss function and the training data;
and inputting the two-dimensional color information of the image to be compressed into the self-coding neural network, and outputting image compression data.
2. The method of claim 1, wherein the predetermined order is set to three orders.
3. The artificial intelligence based image data compression method of claim 1, wherein the performing the gaussian fitting according to the two-dimensional standard deviation of the two-dimensional color information of each pixel point in the training image on a preset order comprises:
and carrying out Gaussian fitting by taking the two-dimensional color information of the pixel points as a center and taking the two-dimensional standard deviation as a model mean value to obtain the Gaussian model.
4. The artificial intelligence based image data compression method of claim 1, wherein the obtaining the boundary two-dimensional color information of the two-dimensional color information according to the two-dimensional standard deviation comprises:
the boundary two-dimensional color information comprises negative boundary two-dimensional color information and positive boundary two-dimensional color information; subtracting the two-dimensional standard deviation from the two-dimensional color information to obtain negative boundary two-dimensional color information; and adding the two-dimensional color information and the standard deviation to obtain positive boundary two-dimensional color information.
5. The artificial intelligence based image data compression method of claim 4, wherein the obtaining a color trend according to the difference between the compressed color information and the boundary two-dimensional color information on the Gaussian model comprises:
obtaining the color tropism according to a color tropism formula; the color tendency formula includes:
wherein, qxiThe color tropism, G, of the ith pixel pointi(yi') is a model value of said compressed color information of said i-th pixel point on said Gaussian model, (X)i，Yi) The two-dimensional color information of the ith pixel point,
the two-dimensional standard deviation of the ith pixel point,
and the model value of the boundary two-dimensional color information of the ith pixel point on the Gaussian model is obtained.
6. The artificial intelligence based image data compression method of claim 1 or 5, wherein the constructing the loss function of the self-coding neural network according to the difference between the first color difference and the second color difference and the color tendency comprises:
the loss function includes:
wherein loss is the loss function, I is the total number of the pixel points, qxiThe color tropism of the ith pixel point, S is the neighborhood range, yiThe two-dimensional color information, m, for the ith pixel pointjThe two-dimensional color information of the jth other pixel point in the neighborhood range, E (y)i,mj) Is said first color difference, yi' is the compressed color information, m ' of the ith pixel point 'jThe compressed color information for the jth other pixel point in the neighborhood, E (y)i′,m′j) Is the second color difference.
7. The method according to claim 1, wherein the two-dimensional color information of the image to be compressed is input into the self-coding neural network, and outputting the compressed image data further comprises
Acquiring gray information of the image to be compressed, inputting the gray information of the image to be compressed and the two-dimensional color information into the self-coding neural network, and outputting composite image compressed data; the composite image compressed data includes the image compressed data and the gradation information.
8. The artificial intelligence based image data compression method of claim 7, wherein outputting the composite image compressed data further comprises:
decoding the composite image compressed data to obtain color information and the gray information; and replacing the gray information with the brightness component information in the color information to obtain a restored image.
9. An artificial intelligence based image data compression system comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111471673.1A CN114170331B (en) | 2021-12-06 | 2021-12-06 | Image data compression method and system based on artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111471673.1A CN114170331B (en) | 2021-12-06 | 2021-12-06 | Image data compression method and system based on artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114170331A true CN114170331A (en) | 2022-03-11 |
| CN114170331B CN114170331B (en) | 2024-09-10 |
Family
ID=80483037
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111471673.1A Active CN114170331B (en) | 2021-12-06 | 2021-12-06 | Image data compression method and system based on artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114170331B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115604477A (en) * | 2022-12-14 | 2023-01-13 | 广州波视信息科技股份有限公司（Cn） | Ultrahigh-definition video distortion optimization coding method |
| CN116228606A (en) * | 2023-05-09 | 2023-06-06 | 南京茂聚智能科技有限公司 | Image optimization processing system based on big data |
| CN118314422A (en) * | 2024-03-13 | 2024-07-09 | 厦门国科安芯科技有限公司 | Loss value calculation method, loss value calculation device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110090950A1 (en) * | 2009-10-15 | 2011-04-21 | General Electric Company | System and method for enhancing data compression using dynamic learning and control |
| CN106254879A (en) * | 2016-08-31 | 2016-12-21 | 广州精点计算机科技有限公司 | A kind of Image Lossy Compression method applying own coding neutral net |
| US20190354811A1 (en) * | 2017-12-07 | 2019-11-21 | Shanghai Cambricon Information Technology Co., Ltd | Image compression method and related device |
| CN111163314A (en) * | 2018-11-07 | 2020-05-15 | 合肥图鸭信息科技有限公司 | Image compression method and system |
| CN112261415A (en) * | 2020-10-23 | 2021-01-22 | 青海民族大学 | Image Compression Coding Method Based on Overfitting Convolutional Autoencoder Network |
-
2021
- 2021-12-06 CN CN202111471673.1A patent/CN114170331B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110090950A1 (en) * | 2009-10-15 | 2011-04-21 | General Electric Company | System and method for enhancing data compression using dynamic learning and control |
| CN106254879A (en) * | 2016-08-31 | 2016-12-21 | 广州精点计算机科技有限公司 | A kind of Image Lossy Compression method applying own coding neutral net |
| US20190354811A1 (en) * | 2017-12-07 | 2019-11-21 | Shanghai Cambricon Information Technology Co., Ltd | Image compression method and related device |
| CN111163314A (en) * | 2018-11-07 | 2020-05-15 | 合肥图鸭信息科技有限公司 | Image compression method and system |
| CN112261415A (en) * | 2020-10-23 | 2021-01-22 | 青海民族大学 | Image Compression Coding Method Based on Overfitting Convolutional Autoencoder Network |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115604477A (en) * | 2022-12-14 | 2023-01-13 | 广州波视信息科技股份有限公司（Cn） | Ultrahigh-definition video distortion optimization coding method |
| CN115604477B (en) * | 2022-12-14 | 2023-03-31 | 广州波视信息科技股份有限公司 | Ultrahigh-definition video distortion optimization coding method |
| CN116228606A (en) * | 2023-05-09 | 2023-06-06 | 南京茂聚智能科技有限公司 | Image optimization processing system based on big data |
| CN118314422A (en) * | 2024-03-13 | 2024-07-09 | 厦门国科安芯科技有限公司 | Loss value calculation method, loss value calculation device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114170331B (en) | 2024-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114170331A (en) | A kind of image data compression method and system based on artificial intelligence | |
| CN107968962B (en) | A video generation method for two non-adjacent images based on deep learning | |
| US20220164934A1 (en) | Image processing method and apparatus, device, video processing method and storage medium | |
| CN110097609B (en) | Sample domain-based refined embroidery texture migration method | |
| CN111145290B (en) | Image colorization method, system and computer readable storage medium | |
| CN111881920B (en) | A network adaptation method for large resolution images and a neural network training device | |
| CN112991371B (en) | An automatic image coloring method and system based on coloring overflow constraints | |
| CN113379604B (en) | Pruning quantization compression method, system and medium for super-resolution network | |
| CN114418121A (en) | Model training method, object processing method and device, electronic device and medium | |
| CN112164005A (en) | Image color correction method, device, equipment and storage medium | |
| CN116309182A (en) | A low-light image enhancement method based on recursive interactive attention | |
| JP2020024524A (en) | A program that replaces all connected layers of a neural network with convolutional layers | |
| CN114648457A (en) | Image enhancement method, apparatus, device and readable storage medium | |
| CN119205493A (en) | A hyperspectral compressed sensing reconstruction method based on deep dual prior progressive fusion | |
| CN113111791B (en) | Image filter conversion network training method and computer readable storage medium | |
| CN112734673A (en) | Low-illumination image enhancement method and system based on multi-expression fusion | |
| CN116977169A (en) | Data processing method, apparatus, device, readable storage medium, and program product | |
| Huang et al. | Edge device-based real-time implementation of CycleGAN for the colorization of infrared video | |
| KR20240064698A (en) | Feature map encoding and decoding method and device | |
| CN114155562A (en) | Gesture recognition method and device | |
| CN120163719A (en) | Image fusion method and system based on two-stage adversarial training and edge perception | |
| CN120070214A (en) | Zero-reference low-illumination image enhancement method based on residual quotient learning | |
| CN118982471A (en) | Underwater image enhancement method, device, storage medium and electronic device | |
| Atsawaraungsuk et al. | Evolutionary circular-ELM for the reduced-reference assessment of perceived image quality | |
| CN114548378A (en) | Neural network generation method and device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |