CN108174218A - Coding and decoding video frame based on study - Google Patents
Coding and decoding video frame based on study Download PDFInfo
- Publication number
- CN108174218A CN108174218A CN201810064012.9A CN201810064012A CN108174218A CN 108174218 A CN108174218 A CN 108174218A CN 201810064012 A CN201810064012 A CN 201810064012A CN 108174218 A CN108174218 A CN 108174218A
- Authority
- CN
- China
- Prior art keywords
- coding
- frame
- space
- time domain
- decoding
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/146—Data rate or code amount at the encoder output
- H04N19/147—Data rate or code amount at the encoder output according to rate distortion criteria
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/189—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding
- H04N19/192—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding the adaptation method, adaptation tool or adaptation type being iterative or recursive
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/70—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by syntax aspects related to video coding, e.g. related to compression standards
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a kind of coding and decoding video frame based on study, including:Space-time domain reconstructs memory, for storing the encoded reconstructing video content with after decoding;Space-time domain predicts network, for using the Space-time domain correlation of reconstructing video content, being modeled by convolutional neural networks and Recognition with Recurrent Neural Network to it, exporting the predicted value of present encoding block;Predicted value subtracts each other to form residual error with original value;Iterative analysis device and iteration synthesizer, step by step to inputting residual error encoding and decoding;Binaryzation device represents the output quantization of iterative analysis device for two-value;Entropy coder carries out entropy coding to the coding output after quantization, obtains output code flow afterwards;Entropy decoder carries out entropy decoding to output code flow, exports give iteration synthesizer afterwards.The coding framework realizes the prediction of Space-time domain by the VoxelCNN (Space-time domain prediction network) based on study, and the control of Video coding rate-distortion optimization is realized with the method that residual error iteration encodes.
Description
Technical field
The present invention relates to video coding and decoding technology field more particularly to a kind of coding and decoding video frames based on study.
Background technology
Existing image/video coding standard is such as:JPEG, H.261, MPEG-2, H.264, H.265, be based on mixing and compile
Code frame.By development for many years, promotion being continuously increased along with complexity of coding efficiency is further compiled in existing mixing
Coding efficiency, which is promoted, under code framework also faces more and more challenges.
But hybrid encoding frame usually realizes the Optimized Coding Based of image/video according to didactic method at present, it is more next
It is more difficult to meet the complexity such as recognition of face instantly, target following, image retrieval, the media application demand of intelligence.
Invention content
The object of the present invention is to provide a kind of coding and decoding video frames based on study, can realize that video encoding rate is distorted
The control of optimization.
The purpose of the present invention is what is be achieved through the following technical solutions:
A kind of coding and decoding video frame based on study, which is characterized in that including:Coding side and decoding end;Wherein encode
End includes:Space-time domain reconstruct memory, Space-time domain prediction network, iterative analysis device, iteration synthesizer, binaryzation device, entropy coder
And entropy decoder;
The Space-time domain reconstructs memory, for storing the encoded reconstructing video content with after decoding;
The Space-time domain predicts network, for using the Space-time domain correlation of reconstructing video content, passing through convolutional Neural net
Network and Recognition with Recurrent Neural Network model it, export the predicted value of present encoding block;
The iterative analysis device, comprising convolutional neural networks and Recognition with Recurrent Neural Network structure, by the pre- survey grid of the Space-time domain
The predicted value of network output, as inputting, exports the compression expression for the residual error with the original residual error to be formed of subtracting each other;
The iteration synthesizer comprising convolutional neural networks and Recognition with Recurrent Neural Network structure, receives entropy decoder decoding production
The compression expression of raw above-mentioned residual error, and the predicted value of the Space-time domain prediction network output is superimposed, form reconstructing video content;
The iterative analysis device and iteration synthesizer, step by step to inputting residual error encoding and decoding, by increase code stream for cost by
Step reduces the distortion level of residual error, realizes the coding of different distortion levels in the case of high-low code flow;
The binaryzation device represents the output quantization of iterative analysis device for two-value;
The entropy coder carries out entropy coding to the coding output after quantization, obtains output code flow afterwards;
The entropy decoder carries out entropy decoding to output code flow, exports give iteration synthesizer afterwards.
As seen from the above technical solution provided by the invention, Space-time domain prediction and residual error iteration coding staff are integrated with
Method is realized the prediction of Space-time domain by the VoxelCNN (Space-time domain prediction network) based on study, and is encoded with residual error iteration
Method realize Video coding rate-distortion optimization control.
Description of the drawings
In order to illustrate the technical solution of the embodiments of the present invention more clearly, required use in being described below to embodiment
Attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description is only some embodiments of the present invention, for this
For the those of ordinary skill in field, without creative efforts, other are can also be obtained according to these attached drawings
Attached drawing.
Fig. 1 is a kind of coding and decoding video block schematic illustration based on study provided in an embodiment of the present invention;
Fig. 2 is the main processes schematic diagram of coding and decoding video frame provided in an embodiment of the present invention;
Fig. 3 is movement Interpolation Process schematic diagram provided in an embodiment of the present invention;
Fig. 4 extends process schematic for movement provided in an embodiment of the present invention.
Specific embodiment
With reference to the attached drawing in the embodiment of the present invention, the technical solution in the embodiment of the present invention is carried out clear, complete
Ground describes, it is clear that described embodiment is only part of the embodiment of the present invention, instead of all the embodiments.Based on this
The embodiment of invention, the every other implementation that those of ordinary skill in the art are obtained without making creative work
Example, belongs to protection scope of the present invention.
The embodiment of the present invention provides a kind of coding and decoding video frame based on study, which mainly wraps
It includes:Coding side and decoding end;As shown in Figure 1, wherein coding side mainly includes:Space-time domain reconstructs memory, Space-time domain predicts network,
Iterative analysis device, iteration synthesizer, binaryzation device, entropy coder and entropy decoder;
The Space-time domain reconstructs memory, for storing the encoded reconstructing video content with after decoding, comprising having solved
Decoded piece of the frame and present frame of code.Encoding-decoding process usually according to the forward direction (P-frame) of video time axis or
Two-way (B-frame) is carried out, and each usual block-by-block of frame is according to sequence encoding and decoding from left to right, from top to bottom.
The Space-time domain prediction network (VoxelCNN), for using the Space-time domain correlation of reconstructing video content, passing through
Convolutional neural networks and Recognition with Recurrent Neural Network model it, export the predicted value of present encoding block;Predicted value and original value phase
Subtract to form residual error, and pass through iterative analysis device and be iterated coding with iteration synthesizer, realize rate-distortion optimization.
The iterative analysis device, comprising convolutional neural networks and Recognition with Recurrent Neural Network structure, by the pre- survey grid of the Space-time domain
The predicted value of network output, as inputting, exports the compression expression for the residual error with the original residual error to be formed of subtracting each other;
The iteration synthesizer comprising convolutional neural networks and Recognition with Recurrent Neural Network structure, receives entropy decoder decoding production
The compression expression of raw above-mentioned residual error, and the predicted value of the Space-time domain prediction network output is superimposed, form reconstructing video content;
The iterative analysis device and iteration synthesizer, step by step to inputting residual error encoding and decoding, by increase code stream for cost by
Step reduces the distortion level of residual error, realizes the coding of different distortion levels in the case of high-low code flow;
The binaryzation device represents the output quantization of iterative analysis device for two-value;
The entropy coder carries out entropy coding to the coding output after quantization, obtains output code flow afterwards;
The entropy decoder after carrying out entropy decoding to output code flow, exports and gives iteration synthesizer.
Entropy coder, entropy decoder can utilize the methods of arithmetic coding/decoding based on context real in the embodiment of the present invention
It is existing, i.e., using arithmetic encoder/decoder as entropy encoder/decoder.
In the embodiment of the present invention, the Space-time domain reconstruct memory, Space-time domain prediction network, iteration synthesizer and entropy decoder
Form the decoder in coding side.
It will be understood by those skilled in the art that because decoding end is only capable of obtaining reconstructing video content rather than original video content,
Therefore coding side includes decoding function to provide reconstructing video content for encoder reference.
In order to make it easy to understand, with reference to specific example as shown in Figure 2 to the main processing in coding and decoding video frame
Process is described in detail.
In the embodiment of the present invention, the predicted value of the Space-time domain prediction network calculations encoding block includes movement synthesis with mixing
Predict two processes.
1st, movement synthesis.
Movement is synthesized comprising interpolation and movement extension is moved, and is two kinds of different coding patterns, and in operation optionally wherein one
Kind mode.
1) movement interpolation is according to adjacent two frame in reconstructing video content to obtain movement locus of object and interpolation is to adjacent two
Between frame, as interpolation frame.As shown in figure 3, movement Interpolation Process is as follows:Enable vx,vy,x,Wherein (vx,vy) represent fortune
Dynamic vector,Represent set of integers.Interpolation frame is denoted asAdjacent two frame is denoted as respectively in reconstructing video contentWithIt is logical
Cross the motion vector (v that the operation of motion compensation that coding block size is m determines an encoding block centered on coordinate (x, y)x,
vy), interpolation frameIn encoding block centered on (x, y)Value byIn withCentered on encoding blockDuplication obtains, and in this manner, can obtain a complete interpolation frameAnd as the output of movement interpolation operation.
2) movement, which is extended to, obtains movement locus of object by the front cross frame of reconstructing video content and extends back, so as to obtain
Obtain an extension frameAs shown in figure 4, movement extension process is as follows:First in front cross frameWithIn, pass through encoding block ruler
The very little operation of motion compensation for m determines the motion vector (v of an encoding block centered on coordinate (x, y)x,vy), extend frameIn encoding block centered on (x, y)Value byIn with (x-vx,y-vy) centered on encoding blockDuplication obtains,
In this manner, a complete extension frame can be obtainedAnd as the output of movement extension operation.
2nd, hybrid predicting.
Hybrid predicting includes convolution and convolution LSTM structures, and (interpolation frame or extension frame being assumed, movement synthesized in Fig. 2
Done in journey movement extension operation, then referred to herein as extension frame), interpolation frame or extend frame front cross frame (With), with
And positioned in present frame above present encoding block with decoded piece of left as inputting, video Space-time domain is believed by study
The modeling of breath generates the predicted value of present encoding block in present frame;By iterative calculation, according to from left to right, from top to bottom
Sequentially, the predicted value of present encoding block can be all generated each time, finally risk totality.
As illustrated in fig. 2, it is assumed that using movement stretched-out coding pattern, then in the case where moving extension mode, will extend two before frame
Frame (With) and in present frame decoded piece of present encoding block upper left side (for each frame according to from upper
Arrive down, sequence from left to right carries out encoding and decoding) as input;In the case where moving interpolation pattern, by the former frame of interpolation frame with after
One frame (With) and decoded piece of present encoding block upper left side is used as input in present frame.Hybrid predicting
By learning the modeling to video space time-domain information, the predicted value of present encoding block is generated;By iterative calculation, according on to
Under, sequence from left to right can all generate the predicted value of present encoding block, finally risk totality each time.The embodiment of the present invention
In, the predicted value of Space-time domain prediction network output subtracts each other to form residual error, and pass through iterative analysis device and synthesize with iteration with original value
Device is iterated coding, and the optimization aim of Space-time domain prediction network is:
Wherein, B is the involved totalframes of optimization, and J is the coding number of blocks that each frame is total in reconstructing video content,
Original value, the predicted value of j-th of encoding block in the i-th frame are corresponded to respectively.
In the embodiment of the present invention, optimization aim is equivalent to loss function, and Space-time domain prediction role of network is to generate predicted value,
And cause this predicted value close to original value.
In the embodiment of the present invention, iterative analysis device is formed with iteration synthesizer comprising the S self-encoding encoders based on convolution
S coding stage, constantly with synthesis to realize variable compression ratio, each stage changes iterative analysis for reconstruction value and desired value
A compression for inputting residual error is generated for analyzer to express, compression expression forms output code flow, iteration point after quantization
The optimization aim of parser and iteration synthesizer is expressed as:
Wherein,For the starting stage (i.e. the 1st stage) input residual error,It represents to input during n-th of stage residual
Difference,Represent the output (i.e. the n stage expresses for inputting the compression of residual error) in n-th of stage.
In the embodiment of the present invention, iterative analysis device and iteration synthesizer are combined optimizations, in formulaActuallyThe output of iterative analysis device, binaryzation device, iteration synthesizer is have passed through, therefore parameter here contains iterative analysis
All parameters in device, iteration synthesizer.
Said program provided in an embodiment of the present invention solves and is difficult to realize that movement is pre- by integration trainingt in neural network
The problems such as survey, proposes VoxelCNN to model the Space-time domain priori of video content simultaneously, and integrated iterative analysis device/synthesizer,
Binaryzation device, entropy encoder/decoder etc. realize the coding and decoding video based on study.In verification experimental verification, in no entropy coding
In the case of device/decoder, the performance of this method has been more than Moving Picture Experts Group-2 encoder, has been reached and H.264 approximate effect.
The foregoing is only a preferred embodiment of the present invention, but protection scope of the present invention be not limited thereto,
Any one skilled in the art is in the technical scope of present disclosure, the change or replacement that can readily occur in,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with the protection model of claims
Subject to enclosing.
Claims (7)
1. a kind of coding and decoding video frame based on study, which is characterized in that including:Coding side and decoding end;Wherein coding side
Including:Space-time domain reconstruct memory, Space-time domain prediction network, iterative analysis device, iteration synthesizer, binaryzation device, entropy coder with
And entropy decoder;
The Space-time domain reconstructs memory, for storing the encoded reconstructing video content with after decoding;
The Space-time domain predicts network, for using the Space-time domain correlation of reconstructing video content, by convolutional neural networks with
And Recognition with Recurrent Neural Network models it, exports the predicted value of present encoding block;
The iterative analysis device, it is comprising convolutional neural networks and Recognition with Recurrent Neural Network structure, Space-time domain prediction network is defeated
The predicted value gone out, as inputting, exports the compression expression for the residual error with the original residual error to be formed of subtracting each other;
The iteration synthesizer comprising convolutional neural networks and Recognition with Recurrent Neural Network structure, receives what entropy decoder decoding generated
The compression expression of above-mentioned residual error, and the predicted value of the Space-time domain prediction network output is superimposed, form reconstructing video content;
The iterative analysis device and iteration synthesizer step by step to inputting residual error encoding and decoding, are gradually subtracted by increasing code stream for cost
The distortion level of few residual error realizes the coding of different distortion levels in the case of high-low code flow;
The binaryzation device represents the output quantization of iterative analysis device for two-value;
The entropy coder carries out entropy coding to the coding output after quantization, obtains output code flow afterwards;
The entropy decoder carries out entropy decoding to output code flow, exports give iteration synthesizer afterwards.
A kind of 2. coding and decoding video frame based on study according to claim 1, which is characterized in that the Space-time domain weight
Structure memory, Space-time domain prediction network, iteration synthesizer form the decoder in coding side with entropy decoder.
3. a kind of coding and decoding video frame based on study according to claim 1, which is characterized in that the Space-time domain is pre-
The predicted value of survey grid network calculation code block includes movement synthesis and two processes of hybrid predicting, wherein:
Movement synthesizes movement interpolation or movement extension operation, and movement interpolation is to be obtained by adjacent two frame of reconstructing video content
Movement locus of object and interpolation are between adjacent two frame, as interpolation frame;Movement is extended to by two before reconstructing video content
Frame obtains movement locus of object and extends back, so as to obtain an extension frame;
Hybrid predicting includes convolution and convolution LSTM structures, interpolation frame is either extended two before frame, interpolation frame or extension frame
Frame and positioned in present frame above present encoding block with decoded piece of left as inputting, by learning to video space
The modeling of time-domain information generates the predicted value of present encoding block in present frame;By iterative calculation, each encoding block is finally obtained
Predicted value.
A kind of 4. coding and decoding video frame based on study according to claim 3, which is characterized in that movement Interpolation Process
It is as follows:Interpolation frame is denoted asAdjacent two frame is denoted as respectively in reconstructing video contentWithIt is m by coding block size
Operation of motion compensation determine the motion vector (v of an encoding block centered on coordinate (x, y)x,vy), interpolation frameIn with
Encoding block centered on (x, y)Value byIn withCentered on encoding blockDuplication obtains, according to
Such method obtains a complete interpolation frame
A kind of 5. coding and decoding video frame based on study according to claim 3, which is characterized in that movement extension process
It is as follows:In reconstructing video content front cross frameWithIn, determine one by the operation of motion compensation that coding block size is m
Motion vector (the v of encoding block centered on coordinate (x, y)x,vy), extend frameIn encoding block centered on (x, y)
Value byIn with (x-vx,y-vy) centered on encoding blockDuplication obtains, and in this manner, obtains one completely
Extend frame
6. a kind of coding and decoding video frame based on study according to claim 1 or 3, which is characterized in that Space-time domain is pre-
The predicted value of survey grid network output subtracts each other to form residual error with original value, and passes through iterative analysis device and be iterated volume with iteration synthesizer
Code, the optimization aim of Space-time domain prediction network are:
Wherein, B is the involved totalframes of optimization, and J is the coding number of blocks that each frame is total in reconstructing video content, Respectively
Correspond to original value, the predicted value of j-th of encoding block in the i-th frame.
7. a kind of coding and decoding video frame based on study according to claim 6, which is characterized in that iterative analysis device with
Iteration synthesizer includes S coding stage of the S self-encoding encoder compositions based on convolution, and reconstruction value and desired value constantly change
With synthesis to realize variable compression ratio, each stage iterative analysis device generates a compaction table for inputting residual error for generation analysis
It reaches, compression expression forms output code flow after quantization, and iterative analyzer and the optimization aim of iteration synthesizer are expressed as:
Wherein,For the starting stage input residual error,Represent the residual error inputted during n-th of stage,Represent n-th of rank
The output of section.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810064012.9A CN108174218B (en) | 2018-01-23 | 2018-01-23 | Video coding and decoding system based on learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810064012.9A CN108174218B (en) | 2018-01-23 | 2018-01-23 | Video coding and decoding system based on learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108174218A true CN108174218A (en) | 2018-06-15 |
CN108174218B CN108174218B (en) | 2020-02-07 |
Family
ID=62515681
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810064012.9A Active CN108174218B (en) | 2018-01-23 | 2018-01-23 | Video coding and decoding system based on learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108174218B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110493596A (en) * | 2019-09-02 | 2019-11-22 | 西北工业大学 | A kind of video coding framework neural network based |
CN111050174A (en) * | 2019-12-27 | 2020-04-21 | 清华大学 | Image compression method, device and system |
CN111222532A (en) * | 2019-10-23 | 2020-06-02 | 西安交通大学 | Edge cloud collaborative deep learning model training method with classification precision maintenance and bandwidth protection |
WO2020107877A1 (en) * | 2018-11-29 | 2020-06-04 | 北京市商汤科技开发有限公司 | Video compression processing method and apparatus, electronic device, and storage medium |
CN111669601A (en) * | 2020-05-21 | 2020-09-15 | 天津大学 | Intelligent multi-domain joint prediction coding method and device for 3D video |
CN111898638A (en) * | 2020-06-29 | 2020-11-06 | 北京大学 | Image processing method, electronic device and medium fusing different visual tasks |
CN113473149A (en) * | 2021-05-14 | 2021-10-01 | 北京邮电大学 | Semantic channel joint coding method and device for wireless image transmission |
WO2022194137A1 (en) * | 2021-03-17 | 2022-09-22 | 华为技术有限公司 | Video image encoding method, video image decoding method and related devices |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1857001A (en) * | 2003-05-20 | 2006-11-01 | Amt先进多媒体科技公司 | Hybrid video compression method |
CN105163121A (en) * | 2015-08-24 | 2015-12-16 | 西安电子科技大学 | Large-compression-ratio satellite remote sensing image compression method based on deep self-encoding network |
CN105430415A (en) * | 2015-12-02 | 2016-03-23 | 宁波大学 | Fast intraframe coding method of 3D-HEVC depth videos |
CN107105278A (en) * | 2017-04-21 | 2017-08-29 | 中国科学技术大学 | The coding and decoding video framework that motion vector is automatically generated |
-
2018
- 2018-01-23 CN CN201810064012.9A patent/CN108174218B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1857001A (en) * | 2003-05-20 | 2006-11-01 | Amt先进多媒体科技公司 | Hybrid video compression method |
CN105163121A (en) * | 2015-08-24 | 2015-12-16 | 西安电子科技大学 | Large-compression-ratio satellite remote sensing image compression method based on deep self-encoding network |
CN105430415A (en) * | 2015-12-02 | 2016-03-23 | 宁波大学 | Fast intraframe coding method of 3D-HEVC depth videos |
CN107105278A (en) * | 2017-04-21 | 2017-08-29 | 中国科学技术大学 | The coding and decoding video framework that motion vector is automatically generated |
Non-Patent Citations (2)
Title |
---|
A.V.D.OORD ET AL.: "Pixel Recurrent Neural Networks", 《INTERNATIONAL CONFERENCE ON MACHINE LEARNING》 * |
FENG JIANG ET AL.: "An End-to-End Compression Framework Based on Convolutional Neural Networks", 《IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020107877A1 (en) * | 2018-11-29 | 2020-06-04 | 北京市商汤科技开发有限公司 | Video compression processing method and apparatus, electronic device, and storage medium |
US11290723B2 (en) | 2018-11-29 | 2022-03-29 | Beijing Sensetime Technology Development Co., Ltd. | Method for video compression processing, electronic device and storage medium |
CN110493596A (en) * | 2019-09-02 | 2019-11-22 | 西北工业大学 | A kind of video coding framework neural network based |
CN111222532A (en) * | 2019-10-23 | 2020-06-02 | 西安交通大学 | Edge cloud collaborative deep learning model training method with classification precision maintenance and bandwidth protection |
CN111222532B (en) * | 2019-10-23 | 2024-04-02 | 西安交通大学 | Training method for edge cloud collaborative deep learning model with classification precision maintenance and bandwidth protection |
CN111050174A (en) * | 2019-12-27 | 2020-04-21 | 清华大学 | Image compression method, device and system |
CN111669601A (en) * | 2020-05-21 | 2020-09-15 | 天津大学 | Intelligent multi-domain joint prediction coding method and device for 3D video |
CN111669601B (en) * | 2020-05-21 | 2022-02-08 | 天津大学 | Intelligent multi-domain joint prediction coding method and device for 3D video |
CN111898638A (en) * | 2020-06-29 | 2020-11-06 | 北京大学 | Image processing method, electronic device and medium fusing different visual tasks |
CN111898638B (en) * | 2020-06-29 | 2022-12-02 | 北京大学 | Image processing method, electronic device and medium fusing different visual tasks |
WO2022194137A1 (en) * | 2021-03-17 | 2022-09-22 | 华为技术有限公司 | Video image encoding method, video image decoding method and related devices |
CN113473149A (en) * | 2021-05-14 | 2021-10-01 | 北京邮电大学 | Semantic channel joint coding method and device for wireless image transmission |
Also Published As
Publication number | Publication date |
---|---|
CN108174218B (en) | 2020-02-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108174218A (en) | Coding and decoding video frame based on study | |
CN107105278A (en) | The coding and decoding video framework that motion vector is automatically generated | |
CN105872558A (en) | Method for performing local motion vector derivation in coding unit | |
CN107835421A (en) | Method and apparatus to Video coding and the method and apparatus to video decoding | |
CN103947211B (en) | Produce the video coding apparatus and method of subregion bit stream | |
CN106534854A (en) | Method of coding and decoding images, coding and decoding device and computer programs corresponding thereto | |
CN110290386B (en) | Low-bit-rate human motion video coding system and method based on generation countermeasure network | |
CN106791848B (en) | Two-Pass code rate control method based on HEVC | |
CN112866694A (en) | Intelligent image compression optimization method combining asymmetric volume block and condition context | |
CN108174204A (en) | A kind of interframe fast schema selection method based on decision tree | |
WO2021262053A1 (en) | Method and system for image compressing and coding with deep learning | |
CN110493596A (en) | A kind of video coding framework neural network based | |
CN103313058B (en) | The HEVC Video coding multimode optimization method realized for chip and system | |
CN115668952B (en) | Method, apparatus and computer readable storage medium for video encoding | |
CN103888770B (en) | A kind of video code conversion system efficiently and adaptively based on data mining | |
CN113132735A (en) | Video coding method based on video frame generation | |
CN117354523A (en) | Image coding, decoding and compressing method for frequency domain feature perception learning | |
CN105556850B (en) | Encoder, decoder and its operation method | |
CN110677644B (en) | Video coding and decoding method and video coding intra-frame predictor | |
CN112770120B (en) | 3D video depth map intra-frame rapid coding method based on depth neural network | |
CN107852493A (en) | Picture decoding method and device for the method for encoding images and device of sample value compensation and for sample value compensation | |
CN111343458B (en) | Sparse gray image coding and decoding method and system based on reconstructed residual | |
CN117093830A (en) | User load data restoration method considering local and global | |
CN111080729A (en) | Method and system for constructing training picture compression network based on Attention mechanism | |
CN106851301A (en) | Dynamic image prediction decoding method, dynamic image prediction decoding device |
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 |