CN108462876A - A kind of video decoding optimization adjusting apparatus and method - Google Patents
A kind of video decoding optimization adjusting apparatus and method Download PDFInfo
- Publication number
- CN108462876A CN108462876A CN201810051581.XA CN201810051581A CN108462876A CN 108462876 A CN108462876 A CN 108462876A CN 201810051581 A CN201810051581 A CN 201810051581A CN 108462876 A CN108462876 A CN 108462876A
- Authority
- CN
- China
- Prior art keywords
- unit
- scene
- circuit
- sent
- video
- 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
- H04N19/20—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using video object coding
-
- 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
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/179—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a scene or a shot
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
A kind of video decoding optimization adjusting apparatus of present invention offer and method, in training mode, video decoding circuit is decoded video source file, and the continuous videos image of tape label after decoding is sent to deep learning circuit, scene and content are identified, and completes training;Under Optimizing Mode, decoded continuous videos image is sent to deep learning circuit by video decoding circuit, scene and content are identified by deep learning circuit, and by recognition result be sent to it is described optimize and revise circuit, it is described optimize and revise circuit video image optimized according to scene and content recognition result be sent to display control unit after adjustment processing and show.The present invention can make Video Decoder can with the current decoding picture material of neural network learning and scene classification, and to classification results make it is different optimize and revise, to reach best decoding effect.
Description
Technical field
The present invention relates to a kind of video decoding optimization adjustment circuit and methods, the video decoding based on video decoding circuit
Afterwards, by deep learning video content, to perhaps the classification of scene and optimize and revise in realizing.
Background technology
It is all that will decode the content come directly to play to user that current video decoding, which plays, cannot be directed to current broadcast
Content is put to optimize and adjust.If Video Decoder can be allow to learn and recognize current decoding picture material and scene
Classification, and different decoding adjustment is made for its scene classification, by the effect and user experience of significant increase video playing.
Invention content
The technical problem to be solved in the present invention is to provide a kind of video decoding optimization adjusting apparatus and method, can make
Obtaining Video Decoder can be with the current decoding picture material of neural network learning and scene classification, and makes difference to classification results
Optimize and revise, to reach best decoding effect.
What apparatus of the present invention were realized in:A kind of video decoding optimization adjusting apparatus, including video decoding circuit, depth
Learning circuit and circuit is optimized and revised, the video decoding circuit, the deep learning circuit and described optimizes and revises circuit
It is sequentially connected;
In training mode, the video decoding circuit is decoded video source file, and by tape label after decoding
Continuous videos image is sent to deep learning circuit, and the deep learning circuit is different by using artwork and frame difference figure
Convolution kernel generates scene and content recognition as a result, and the video tab and the scene and content recognition brought according to video source
As a result iteration is trained to train until completing;
Under Optimizing Mode, the video decoding circuit is negative to be decoded video source file, and will be decoded continuous
Video image is sent to deep learning circuit, and the deep learning circuit to artwork and frame difference figure by using different convolution
Core generate scene and content recognition as a result, and by recognition result be sent to it is described optimize and revise circuit, it is described to optimize and revise circuit root
It is sent to display control unit after optimizing adjustment processing to video image according to scene and content recognition result.
What the method for the present invention was realized in:A kind of video decoding optimization method of adjustment, including training process and optimized
Journey;
The training process is:Video source file is decoded by video decoding circuit, and will decoding after tape label
Continuous videos image be sent to deep learning circuit, the deep learning circuit pass through artwork and frame difference figure are used it is different
Convolution kernel generate scene and content recognition as a result, and the video tab brought according to video source know with the scene and content
Other result is trained iteration and is trained until completing;
The optimization process is:Video source file is decoded by the way that video decoding circuit is negative, and by decoded company
Continuous video image is sent to deep learning circuit, and the deep learning circuit to artwork and frame difference figure by using different volumes
Product core generate scene and content recognition as a result, and by recognition result be sent to it is described optimize and revise circuit, it is described to optimize and revise circuit
It is sent to display control unit after optimizing adjustment processing to video image according to scene and content recognition result.
The invention has the advantages that:The present invention makes can be with the current solution of neural network learning during video is decoded
Code picture material and scene classification, and to classification results make it is different optimize and revise, to reach best decoding effect.
Description of the drawings
The present invention is further illustrated in conjunction with the embodiments with reference to the accompanying drawings.
Fig. 1 is the structure and execution flow chart of video decoding optimization adjusting apparatus of the present invention.
Fig. 2 is the structure and execution flow chart of deep learning circuit in the present invention.
Specific implementation mode
Refering to Figure 1, what apparatus of the present invention were realized in:A kind of video decoding optimization adjusting apparatus, including regard
Frequency decoding circuit, deep learning circuit and circuit is optimized and revised, the video decoding circuit, the deep learning circuit and institute
It states and optimizes and revises circuit and be sequentially connected;
In training mode, the video decoding circuit is decoded video source file, and by tape label after decoding
Continuous videos image is sent to deep learning circuit, and the deep learning circuit is different by using artwork and frame difference figure
Convolution kernel generates scene and content recognition as a result, and the video tab and the scene and content recognition brought according to video source
As a result iteration is trained to train until completing;
Under Optimizing Mode, the video decoding circuit is negative to be decoded video source file, and will be decoded continuous
Video image is sent to deep learning circuit, and the deep learning circuit to artwork and frame difference figure by using different convolution
Core generate scene and content recognition as a result, and by recognition result be sent to it is described optimize and revise circuit, it is described to optimize and revise circuit root
It is sent to display control unit after optimizing adjustment processing to video image according to scene and content recognition result.
The present invention also provides a kind of video decoding optimization methods of adjustment, including training process and optimization process;
The training process is:Video source file is decoded by video decoding circuit, and will decoding after tape label
Continuous videos image be sent to deep learning circuit, the deep learning circuit pass through artwork and frame difference figure are used it is different
Convolution kernel generate scene and content recognition as a result, and the video tab brought according to video source know with the scene and content
Other result is trained iteration and is trained until completing;
The optimization process is:Video source file is decoded by the way that video decoding circuit is negative, and by decoded company
Continuous video image is sent to deep learning circuit, and the deep learning circuit to artwork and frame difference figure by using different volumes
Product core generate scene and content recognition as a result, and by recognition result be sent to it is described optimize and revise circuit, it is described to optimize and revise circuit
It is sent to display control unit after optimizing adjustment processing to video image according to scene and content recognition result.
Wherein, the deep learning circuit includes frame buffer unit, frame difference computing unit, artwork neuron input list
Member, inter-frame information input neuron elements and CNN convolutional neural networks units;The frame buffer unit respectively with the video
Decoding circuit, the frame difference computing unit and the artwork neuron input unit;The CNN convolutional neural networks list
Member respectively with the video decoding circuit, artwork neuron input unit, inter-frame information input neuron elements and described excellent
Change adjustment circuit connection.
The frame buffer unit is responsible for storing present frame and previous frame image, and present image is sent to described optimize and revise
Circuit and artwork neuron input unit, frame difference computing unit is sent to by the image of present frame and former frame;
The frame difference computing unit is responsible for calculating the difference of each pixel position between two frames, i.e. frame-to-frame differences
Then frame difference is sent to the inter-frame information and inputs neuron elements by value;
The original graph content input neuron circuit is responsible for CNN volumes of composition with inter-frame information input neuron elements
The neuron input of product neural network;
The CNN convolutional neural networks receive the artwork neuron input unit and input neuron with the inter-frame information
After the input of unit, in training mode, the video tab and the scene and content recognition result brought according to video source
Iteration is trained to train until completing.
It please refers to shown in Fig. 2, the deep learning circuit further includes parameter storage unit, and the parameter storage unit is for depositing
Store up network weight and threshold value, interframe convolution kernel and common convolution kernel;The parameter storage unit can be a unit, can also divide
At three units, storage network weight as shown in Figure 2 and threshold value storage unit, interframe convolution kernel storage unit, common volume
Product core storage unit.
The CNN convolutional neural networks unit further comprises neural network computing unit, parameter initialization unit, convolution
Biasing access unit, convolution kernel access unit, weights access unit, error calculation unit and backpropagation write back unit;Institute
It states neural network computing unit and is separately connected the artwork neuron input unit and inter-frame information input neuron elements;
The parameter initialization unit connects the parameter storage unit;The neural network computing unit is also biased by the convolution
Access unit, convolution kernel access unit, weights access unit connect the parameter storage unit;The neural network computing unit
It passes sequentially through the error calculation unit and backpropagation write back unit connects the parameter storage unit.
In the training process, the CNN convolutional neural networks unit carries out following process:
(1) it initializes:Start train when, the parameter initialization unit according to preset initialization algorithm to convolution kernel,
Three weights, convolution bias parameters are initialized, and initial value is obtained;
(2) it fetches:After the completion of initialization, the original graph content input neuron circuit and inter-frame information input god
Artwork and frame difference figure are sent to neural network computing unit by neuron input data through first unit;Meanwhile the volume
The convolution bias that product biasing access unit reads each layer network from the parameter storage unit is sent to the neural network computing
Unit;All convolution kernel values that the convolution kernel access unit reads each layer network from the parameter storage unit are sent to nerve net
Network arithmetic element;The weights that the weights access unit reads each layer network from the parameter storage unit are sent to neural network fortune
Calculate unit;Wherein, when the parameter storage unit is divided into storage network weight and threshold value storage unit, the storage of interframe convolution kernel are single
When first, common three units of convolution kernel storage unit, convolution kernel access unit be from DDR interframe convolution kernel storage unit and
The use of neural network computing unit is given in common convolution kernel storage unit access, and wherein interframe convolution kernel gives frame difference neuron
The frame difference figure of input uses, and common convolution kernel is used to the original graph that artwork neuron inputs;
(3) operation:After completing all access, the neural network computing unit starts to be transported according to the initial value
It calculates, and obtains operation result, operation result is then sent to the error calculation unit;The error calculation unit is according to calculating
As a result error calculation is carried out with expected results, and the error amount for calculating gained is sent to the backpropagation write back unit;It is described
Backpropagation write back unit calculates the updated value of convolution kernel, three weights, convolution bias parameters according to error amount, then will
Updated value is written back to the correspondence parameter position of DDR;
After completing a wheel training, constantly repeatedly step (2) and (3) until the training completion of all video sources, and reach default
Pattern-recognition accuracy.
Again as shown in Figure 1, the circuit of optimizing and revising includes scene judging unit, content judging unit, post processing of image
Unit and subtitle superposition unit;The scene judging unit and the content judging unit are connected to the CNN convolutional Neurals
Network element, the scene judging unit are also connected with post processing of image unit, and the subtitle superposition unit connects the content and sentences
Order member and the frame buffer unit.
The circuit of optimizing and revising further includes that scene corresponds to post-treatment parameters storage unit, scene corresponds to display parameters storage
Unit and content correspond to subtitle storage unit, and the scene corresponds to after post-treatment parameters storage unit is connected to described image
Unit is managed, the scene corresponds to display parameters storage unit and connects the scene judging unit, and the content corresponds to subtitle storage
Unit connects the subtitle superposition unit.
The scene judging unit be responsible for the scene generated according to convolutional neural networks judgement result be sent to described image after
Processing unit;
The content judging unit is responsible for the content generated according to convolutional neural networks and judges that result is sent to the subtitle and folds
Add unit;
After described image post-processing unit receives scene information, the corresponding post-treatment parameters storage unit of scene is read, it is right
Video image is handled, for example scene is night or dark image, then integrally improves picture contrast;Such as when judging certainly
So scene, then improve image chroma and saturation degree etc.;Then the image after post-processing is sent to display control unit;
The subtitle superposition unit is responsible for judging that result reads corresponding subtitle according to content, such as violence and pornographic interior
Hold, subtitle can add automatically " it is unsuitable for children, ask children to avoid " etc. subtitles;
To video it is decoded during can with the current decoding picture material of neural network learning and scene classification, and
To classification results make it is different optimize and revise, to reach best decoding effect.
Although specific embodiments of the present invention have been described above, those familiar with the art should manage
Solution, we are merely exemplary described specific embodiment, rather than for the restriction to the scope of the present invention, it is familiar with this
The technical staff in field modification and variation equivalent made by the spirit according to the present invention, should all cover the present invention's
In scope of the claimed protection.
Claims (10)
1. a kind of video decoding optimization adjusting apparatus, it is characterised in that:Including video decoding circuit, deep learning circuit and excellent
Change adjustment circuit, the video decoding circuit, the deep learning circuit and the circuit of optimizing and revising are sequentially connected;
In training mode, the video decoding circuit is decoded video source file, and by after decoding tape label it is continuous
Video image is sent to deep learning circuit, and the deep learning circuit to artwork and frame difference figure by using different convolution
Core generates scene and content recognition as a result, and the video tab and the scene and content recognition result brought according to video source
Iteration is trained to train until completing;
Under Optimizing Mode, the video decoding circuit is decoded video source file, and by decoded continuous videos figure
As being sent to deep learning circuit, the deep learning circuit is by generating artwork and frame difference figure using different convolution kernels
Scene and content recognition as a result, and by recognition result be sent to it is described optimize and revise circuit, the circuit of optimizing and revising is according to scene
It is sent to display control unit after optimizing adjustment processing to video image with content recognition result.
2. a kind of video decoding optimization adjusting apparatus according to claim 1, it is characterised in that:The deep learning circuit
Including frame buffer unit, frame difference computing unit, artwork neuron input unit, inter-frame information input neuron elements and
CNN convolutional neural networks units;The frame buffer unit respectively with the video decoding circuit, the frame difference computing unit
And the artwork neuron input unit;The CNN convolutional neural networks unit respectively with the video decoding circuit, artwork
Neuron input unit, inter-frame information input neuron elements and the circuit of optimizing and revising connect.
3. a kind of video decoding optimization adjusting apparatus according to claim 2, it is characterised in that:The deep learning circuit
Further include parameter storage unit, the parameter storage unit is for storing network weight and threshold value, interframe convolution kernel and common convolution
Core;
The CNN convolutional neural networks unit further comprises neural network computing unit, parameter initialization unit, convolution biasing
Access unit, convolution kernel access unit, weights access unit, error calculation unit and backpropagation write back unit;The god
It is separately connected the artwork neuron input unit through network operations unit and the inter-frame information inputs neuron elements;It is described
Parameter initialization unit connects the parameter storage unit;The neural network computing unit is also biased by the convolution and is fetched
Unit, convolution kernel access unit, weights access unit connect the parameter storage unit;The neural network computing unit is successively
The parameter storage unit is connected by the error calculation unit and backpropagation write back unit.
4. a kind of video decoding optimization adjusting apparatus according to claim 2, it is characterised in that:It is described to optimize and revise circuit
Including scene judging unit, content judging unit, post processing of image unit and subtitle superposition unit;The scene judging unit and
The content judging unit is connected to the CNN convolutional neural networks unit, after the scene judging unit is also connected with image
Processing unit, the subtitle superposition unit connect the content judging unit and the frame buffer unit.
5. a kind of video decoding optimization adjusting apparatus according to claim 4, it is characterised in that:It is described to optimize and revise circuit
Further include that scene corresponds to post-treatment parameters storage unit, scene corresponds to display parameters storage unit and content corresponds to subtitle storage
Unit, the scene correspond to post-treatment parameters storage unit and are connected to described image post-processing unit, and the scene corresponds to display
Parameter storage unit connects the scene judging unit, and the content corresponds to subtitle storage unit and connects the subtitle superposition list
Member.
6. a kind of video decoding optimization method of adjustment, it is characterised in that:Including training process and optimization process;
The training process is:Video source file is decoded by video decoding circuit, and by the company of tape label after decoding
Continuous video image is sent to deep learning circuit, and the deep learning circuit to artwork and frame difference figure by using different volumes
Product core generates scene and content recognition as a result, and the video tab and the scene and content recognition knot brought according to video source
Fruit is trained iteration and is trained until completing;
The optimization process is:Video source file is decoded by the way that video decoding circuit is negative, and is continuously regarded decoded
Frequency image is sent to deep learning circuit, and the deep learning circuit to artwork and frame difference figure by using different convolution kernels
Generate scene and content recognition as a result, and by recognition result be sent to it is described optimize and revise circuit, it is described optimize and revise circuit according to
Scene and content recognition result are sent to display control unit after optimizing adjustment processing to video image.
7. a kind of video decoding optimization method of adjustment according to claim 6, it is characterised in that:The deep learning circuit
Including frame buffer unit, frame difference computing unit, artwork neuron input unit, inter-frame information input neuron elements and
CNN convolutional neural networks units;
The frame buffer unit is responsible for storing present frame and previous frame image, and present image is sent to and described optimizes and revises circuit
With artwork neuron input unit, the image of present frame and former frame is sent to frame difference computing unit;
The frame difference computing unit is responsible for calculating the difference of each pixel position between two frames, i.e. frame difference, so
Frame difference is sent to the inter-frame information afterwards and inputs neuron elements;
The original graph content input neuron circuit is responsible for forming CNN convolution god with inter-frame information input neuron elements
Neuron input through network;
The CNN convolutional neural networks receive the artwork neuron input unit and input neuron elements with the inter-frame information
Input after, in training mode, the video tab brought according to video source is carried out with the scene and content recognition result
Training iteration is trained until completing.
8. a kind of video decoding optimization method of adjustment according to claim 7, it is characterised in that:The deep learning circuit
Further include that parameter storage unit and scene correspond to post-treatment parameters storage unit, the parameter storage unit is for storing network weight
Weight and threshold value, interframe convolution kernel and common convolution kernel;The CNN convolutional neural networks unit further comprises neural network computing
Unit, parameter initialization unit, convolution biasing access unit, convolution kernel access unit, weights access unit, error calculation unit
And backpropagation write back unit;In the training process, the CNN convolutional neural networks unit carries out following process:
(1) it initializes:Start train when, the parameter initialization unit according to preset initialization algorithm to convolution kernel, weights,
Three parameters of convolution bias are initialized, and initial value is obtained;
(2) it fetches:After the completion of initialization, the original graph content input neuron circuit inputs neuron with the inter-frame information
Artwork and frame difference figure are sent to neural network computing unit by unit by neuron input data;Meanwhile the convolution is inclined
It sets access unit and reads the convolution bias of each layer network from the parameter storage unit and be sent to the neural network computing unit;
All convolution kernel values that the convolution kernel access unit reads each layer network from the parameter storage unit are sent to neural network fortune
Calculate unit;The weights that the weights access unit reads each layer network from the parameter storage unit are sent to neural network computing list
Member;
(3) operation:After completing all access, the neural network computing unit starts to carry out operation according to the initial value, and
Operation result is obtained, operation result is then sent to the error calculation unit;The error calculation unit is according to result of calculation
Error calculation is carried out with expected results, and the error amount for calculating gained is sent to the backpropagation write back unit;It is described reversed
The updated value that write back unit calculates convolution kernel, three weights, convolution bias parameters according to error amount is propagated, it then will update
Value is written back to the correspondence parameter position of DDR;
After completing a wheel training, constantly repeatedly step (2) and (3) until the training completion of all video sources, and reach preset mould
Formula recognition correct rate.
9. a kind of video decoding optimization method of adjustment according to claim 7, it is characterised in that:It is described to optimize and revise circuit
Including scene judging unit, content judging unit, post processing of image unit and subtitle superposition unit;The scene judging unit and
The content judging unit is connected to the CNN convolutional neural networks unit, after the scene judging unit is also connected with image
Processing unit, the subtitle superposition unit connect the content judging unit and the frame buffer unit;
The scene judgement result that the scene judging unit is generated according to convolutional neural networks is sent to described image post-processing unit;
After described image post-processing unit receives scene information, the corresponding post-treatment parameters storage unit of scene is read, to video image
Display control unit is sent to after being handled;
The content judgement result that the content judging unit is generated according to convolutional neural networks is sent to the subtitle superposition unit;Institute
It states subtitle superposition unit and judges that result reads corresponding subtitle according to content.
10. a kind of video decoding optimization method of adjustment according to claim 9, it is characterised in that:It is described to optimize and revise electricity
Road further includes that scene corresponds to post-treatment parameters storage unit, scene corresponds to display parameters storage unit and content corresponds to subtitle and deposits
Storage unit, the scene correspond to post-treatment parameters storage unit and are connected to described image post-processing unit, and the scene corresponds to aobvious
Show that parameter storage unit connects the scene judging unit, the content corresponds to subtitle storage unit and connects the subtitle superposition list
Member.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810051581.XA CN108462876B (en) | 2018-01-19 | 2018-01-19 | Video decoding optimization adjustment device and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810051581.XA CN108462876B (en) | 2018-01-19 | 2018-01-19 | Video decoding optimization adjustment device and method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108462876A true CN108462876A (en) | 2018-08-28 |
CN108462876B CN108462876B (en) | 2021-01-26 |
Family
ID=63220994
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810051581.XA Active CN108462876B (en) | 2018-01-19 | 2018-01-19 | Video decoding optimization adjustment device and method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108462876B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109840590A (en) * | 2019-01-31 | 2019-06-04 | 福州瑞芯微电子股份有限公司 | A kind of scene classification circuit framework neural network based and method |
CN110458051A (en) * | 2019-07-25 | 2019-11-15 | 中移(杭州)信息技术有限公司 | A kind of method, apparatus, server and the readable storage medium storing program for executing of equipment control |
CN110493639A (en) * | 2019-10-21 | 2019-11-22 | 南京创维信息技术研究院有限公司 | A kind of method and system of adjust automatically sound and image model based on scene Recognition |
CN111031392A (en) * | 2019-12-23 | 2020-04-17 | 广州视源电子科技股份有限公司 | Media file playing method, system, device, storage medium and processor |
CN112631120A (en) * | 2019-10-09 | 2021-04-09 | Oppo广东移动通信有限公司 | PID control method, device and video coding and decoding system |
CN112929703A (en) * | 2019-12-06 | 2021-06-08 | 上海海思技术有限公司 | Method and device for processing code stream data |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102014295A (en) * | 2010-11-19 | 2011-04-13 | 嘉兴学院 | Network sensitive video detection method |
CN102158699A (en) * | 2011-03-28 | 2011-08-17 | 广州市聚晖电子科技有限公司 | Embedded video compression coding system with image enhancement function |
CN102254224A (en) * | 2011-07-06 | 2011-11-23 | 无锡泛太科技有限公司 | Internet of things electric automobile charging station system based on image identification of rough set neural network |
CN104202604A (en) * | 2014-08-14 | 2014-12-10 | 腾讯科技(深圳)有限公司 | Video enhancing method and device |
CN104299196A (en) * | 2014-10-11 | 2015-01-21 | 京东方科技集团股份有限公司 | Image processing device and method and display device |
CN104813674A (en) * | 2012-10-01 | 2015-07-29 | 谷歌公司 | System and method for optimizing videos |
CN104881675A (en) * | 2015-05-04 | 2015-09-02 | 北京奇艺世纪科技有限公司 | Video scene identification method and apparatus |
CN104933680A (en) * | 2015-03-13 | 2015-09-23 | 哈尔滨工程大学 | Intelligent unmanned surface vessel visual system video rapid sea fog removing method |
CN105302872A (en) * | 2015-09-30 | 2016-02-03 | 努比亚技术有限公司 | Image processing device and method |
US20170148487A1 (en) * | 2015-11-25 | 2017-05-25 | Dell Products L.P. | Video Manipulation for Privacy Enhancement |
US20170238055A1 (en) * | 2014-02-28 | 2017-08-17 | Second Spectrum, Inc. | Methods and systems of spatiotemporal pattern recognition for video content development |
CN107590443A (en) * | 2017-08-23 | 2018-01-16 | 上海交通大学 | Limiter stage live video automatic testing method and system based on the study of depth residual error |
-
2018
- 2018-01-19 CN CN201810051581.XA patent/CN108462876B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102014295A (en) * | 2010-11-19 | 2011-04-13 | 嘉兴学院 | Network sensitive video detection method |
CN102158699A (en) * | 2011-03-28 | 2011-08-17 | 广州市聚晖电子科技有限公司 | Embedded video compression coding system with image enhancement function |
CN102254224A (en) * | 2011-07-06 | 2011-11-23 | 无锡泛太科技有限公司 | Internet of things electric automobile charging station system based on image identification of rough set neural network |
CN104813674A (en) * | 2012-10-01 | 2015-07-29 | 谷歌公司 | System and method for optimizing videos |
US20170238055A1 (en) * | 2014-02-28 | 2017-08-17 | Second Spectrum, Inc. | Methods and systems of spatiotemporal pattern recognition for video content development |
CN104202604A (en) * | 2014-08-14 | 2014-12-10 | 腾讯科技(深圳)有限公司 | Video enhancing method and device |
CN104299196A (en) * | 2014-10-11 | 2015-01-21 | 京东方科技集团股份有限公司 | Image processing device and method and display device |
CN104933680A (en) * | 2015-03-13 | 2015-09-23 | 哈尔滨工程大学 | Intelligent unmanned surface vessel visual system video rapid sea fog removing method |
CN104881675A (en) * | 2015-05-04 | 2015-09-02 | 北京奇艺世纪科技有限公司 | Video scene identification method and apparatus |
CN105302872A (en) * | 2015-09-30 | 2016-02-03 | 努比亚技术有限公司 | Image processing device and method |
US20170148487A1 (en) * | 2015-11-25 | 2017-05-25 | Dell Products L.P. | Video Manipulation for Privacy Enhancement |
CN107590443A (en) * | 2017-08-23 | 2018-01-16 | 上海交通大学 | Limiter stage live video automatic testing method and system based on the study of depth residual error |
Non-Patent Citations (1)
Title |
---|
MAURICIO PEREZ等: "Video pornography detection through deep learning techniques and motion", 《NEUROCOMPUTING》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109840590A (en) * | 2019-01-31 | 2019-06-04 | 福州瑞芯微电子股份有限公司 | A kind of scene classification circuit framework neural network based and method |
CN110458051A (en) * | 2019-07-25 | 2019-11-15 | 中移(杭州)信息技术有限公司 | A kind of method, apparatus, server and the readable storage medium storing program for executing of equipment control |
CN112631120A (en) * | 2019-10-09 | 2021-04-09 | Oppo广东移动通信有限公司 | PID control method, device and video coding and decoding system |
CN112631120B (en) * | 2019-10-09 | 2022-05-17 | Oppo广东移动通信有限公司 | PID control method, device and video coding and decoding system |
CN110493639A (en) * | 2019-10-21 | 2019-11-22 | 南京创维信息技术研究院有限公司 | A kind of method and system of adjust automatically sound and image model based on scene Recognition |
WO2021077663A1 (en) * | 2019-10-21 | 2021-04-29 | 南京创维信息技术研究院有限公司 | Method and system for automatically adjusting sound and image modes on basis of scene recognition |
CN112929703A (en) * | 2019-12-06 | 2021-06-08 | 上海海思技术有限公司 | Method and device for processing code stream data |
CN111031392A (en) * | 2019-12-23 | 2020-04-17 | 广州视源电子科技股份有限公司 | Media file playing method, system, device, storage medium and processor |
Also Published As
Publication number | Publication date |
---|---|
CN108462876B (en) | 2021-01-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108462876A (en) | A kind of video decoding optimization adjusting apparatus and method | |
Xu et al. | Generative low-bitwidth data free quantization | |
Yang et al. | Collaborative video object segmentation by foreground-background integration | |
Nam et al. | Batch-instance normalization for adaptively style-invariant neural networks | |
Park et al. | Meta-tracker: Fast and robust online adaptation for visual object trackers | |
Qiu et al. | Learning spatio-temporal representation with pseudo-3d residual networks | |
CN111128137B (en) | Training method and device for acoustic model, computer equipment and storage medium | |
Oh et al. | Fast user-guided video object segmentation by interaction-and-propagation networks | |
CN108763444B (en) | Method for solving video question-answering by using layered coding decoder network mechanism | |
CN107391646A (en) | A kind of Semantic features extraction method and device of video image | |
CN110555881A (en) | Visual SLAM testing method based on convolutional neural network | |
CN109886343B (en) | Image classification method and device, equipment and storage medium | |
US20230169325A1 (en) | Learning energy-based model with variational auto-encoder as amortized sampler | |
CN115578680B (en) | Video understanding method | |
WO2015011521A1 (en) | An incremental learner via an adaptive mixture of weak learners distributed on a non-rigid binary tree | |
US20230131850A1 (en) | Production method of multimedia work, apparatus, and computer-readable storage medium | |
Gauthier et al. | Learning to prove with tactics | |
Yang et al. | Sample efficient reinforcement learning method via high efficient episodic memory | |
CN117710533B (en) | Music conditional dance animation generation method based on diffusion model | |
CN114137967B (en) | Driving behavior decision method based on multi-network joint learning | |
CN110334355B (en) | Relation extraction method, system and related components | |
CN116246214A (en) | Audio-visual event positioning method, model training method, device, equipment and medium | |
CN110505520B (en) | Information recommendation method and system, medium and electronic device | |
US11875809B2 (en) | Speech denoising via discrete representation learning | |
Ding et al. | Video frame interpolation via local lightweight bidirectional encoding with channel attention cascade |
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 | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 350000 building 18, 89 software Avenue, Gulou District, Fuzhou City, Fujian Province Applicant after: Ruixin Microelectronics Co., Ltd Address before: 350000 building 18, 89 software Avenue, Gulou District, Fuzhou City, Fujian Province Applicant before: Fuzhou Rockchips Electronics Co.,Ltd. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |