CN107197260A - Video coding post-filter method based on convolutional neural networks - Google Patents
Video coding post-filter method based on convolutional neural networks Download PDFInfo
- Publication number
- CN107197260A CN107197260A CN201710439132.8A CN201710439132A CN107197260A CN 107197260 A CN107197260 A CN 107197260A CN 201710439132 A CN201710439132 A CN 201710439132A CN 107197260 A CN107197260 A CN 107197260A
- Authority
- CN
- China
- Prior art keywords
- frame
- video
- neural networks
- convolutional neural
- post
- 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
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/117—Filters, e.g. for pre-processing or post-processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/124—Quantisation
-
- 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
- H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
Abstract
Video coding post-filter method based on convolutional neural networks, including convolutional neural networks model training step and filter step, training step include：Set the quantization parameter of video compress to carry out coding compression for 20 to 51 pairs of original videos, obtain compressing video；All videos are carried out with frame and extracts the frame pair for obtaining multiple compressed video frame original video frames；It will extract and obtain frame to being multiple groups by the different demarcation of frame type and quantization parameter；Convolutional neural networks framework and initialization network parameter are built, neutral net is trained respectively using the group of foregoing division, obtains corresponding to different quantization parameters and multiple neural network models of frame type.Filter step includes：Obtained multiple neural network models are embedded in the post-filtering link of video encoder；The coding compression foregoing to pending original video execution and frame extract and obtain pending frame pair, and the quantization parameter and the corresponding neural network model of frame type selecting according to pending frame pair are filtered processing.
Description
Technical field
The present invention relates to computer vision and field of video encoding, and in particular to a kind of video based on convolutional neural networks
Encode post-filter method.
Background technology
With the development of science and technology and enriching for various video display apparatus, video has been increasingly becoming can not in people's life
The part lacked, highly important effect is played in every field.The past video resolution that witnessed in decades is with showing
Show the great development of device screen, and the video council of ultrahigh resolution produces huge data volume, this can be produced to the network bandwidth
Greatly bear.Accordingly, it would be desirable to which efficient Video coding and transmission technology ensure that user watches the experience of video, while to the greatest extent may be used
The data volume of energy ground reduction video, is network bandwidth Reduction of Students' Study Load.In consideration of it, researcher is in decades, constantly research is efficiently regarded
Frequency coding method.Video coding technique, mainly reduced by removing the redundancy in video the data volume of video with up to
To effectively storing and transmitting multitude of video data, it is intended on the premise of keeping former video quality as far as possible, with lower code check
Compress video.
However, current video encoding standard is mainly all based on the hybrid video coding technology of block, in this encoder block
It is block-based to predict within the frame/frames, change and slightly quantify all cause the decline of video quality, especially in low code under frame
In the case of rate.Therefore, the distortion reduced in Video coding turns into one of study hotspot of current video coding field.Although,
Some algorithms are also taken in current video coding standards to reduce blocking effect and improve subjective quality, but its effect is also not
Enough ideals, the video after processing still suffers from obvious blocking effect and edge blurry, and loss in detail is still than more serious.
The content of the invention
It is a primary object of the present invention to using strong capability of fitting of the convolutional neural networks to nonlinear transformation, propose a kind of
Video coding post-filter method based on convolutional neural networks, this method, which is set up, damages frame of video reflecting to lossless video frame
Penetrate, so that the inverse transformation of Distortion course in Video coding is approximately obtained, to reach the purpose for reducing distortion.
The present invention is as follows for the technical scheme provided up to above-mentioned purpose：
A kind of Video coding post-filter method based on convolutional neural networks, includes the training of convolutional neural networks model
Step and post-filtering process step, wherein：
The training step includes S1 to S4：
S1, the quantization parameter of setting video compress are 20 to 51, carry out coding compression to original video, obtain compression and regard
Frequently；
S2, the extraction that frame is carried out to the compression video and the original video, obtain multiple frames pair, each frame pair
Include a compressed video frame and an original video frame；
S3, the frame for extracting step S2 are multiple groups to the different demarcation according to frame type and quantization parameter；
S4, convolutional neural networks framework and initialization network parameter are built, the group divided using step S3 is respectively to nerve
Network is trained, and obtains corresponding to different quantization parameters and multiple neural network models of frame type；
The post-filtering process step includes S5 and S6：
S5, the post-filtering link that multiple neural network models that step S4 is obtained are embedded in video encoder；
S6, step S1 and S2 are performed to pending original video obtain pending frame pair, and according to pending frame pair
Quantization parameter and the corresponding neural network model of frame type selecting are filtered processing.
The difference of quantization parameter and frame type causes their nature of distortion also different, and the present invention is in training and actual treatment
During be all the frame pair for extracting original video and compression video composition, set up with this and damage frame of video and lossless video frame
Mapping；And post-filtering in encoder is embedded into by the neural network model trained for different quantization parameters and frame type
Link is filtered processing, takes into full account the influence of quantization parameter and frame type to distortion level, first using the method for the present invention
Differentiate the quantization parameter and frame type of frame, reselection quantization parameter and all corresponding neural network model of frame type are carried out
Filtering, so as to effectively suppress distortion.
Brief description of the drawings
Fig. 1 is the flow chart for the Video coding post-filter method based on convolutional neural networks that the present invention is provided；
Fig. 2 is convolutional neural networks frame diagram；
Fig. 3 is the training process schematic diagram of convolutional neural networks.
Embodiment
The invention will be further described with specific embodiment below in conjunction with the accompanying drawings.
The present invention is using the powerful nonlinear fitting ability of convolutional neural networks, with reference to the characteristic of Video coding, fully divides
In analysis Video coding compression process the reason for cause distortion, it is proposed that based on convolutional neural networks and be relevant to quantization parameter and frame
The Video coding post-filter method of type, sets up the mapping damaged between frame of video and lossless video frame, and training obtains difference
Convolutional neural networks model under quantization parameter (abbreviation QP) and different frame type, for the post-filtering link in encoder
Processing is filtered for the frame of different quantization parameters and frame type, can effectively suppress distortion.
In mixed video coding framework, frame type mainly has I frames, P frames and B frames, wherein：
I frames use intra prediction mode, also referred to as intracoded frame (intra picture), and I frames are typically every
The start frame of one picture group (GOP), its information independent of front and rear frame, therefore shuffle can be used as in broadcasting
Point of penetration.For intra prediction mode, in decoding end, it is the block and the block of top on the adjacent left side based on decoding, uses
Algorithm carries out the prediction of current block, and then along with residual error data carrys out the raw information of reconstruction image, the residual error data is by solving
Code value and predicted value are subtracted each other and obtained；In coding side, it is necessary to using identical infra-frame prediction, because predicted value and raw image data
It is more similar, so the energy for the remaining data to be transmitted is low-down, so as to realize the compression of code stream.
P frames and B frames use inter-frame forecast mode, P frames are forward predicted frame (predictive-frame), with reference to above
I frames or B frames remove temporal redundancy, realize Efficient Compression.B frames are bi-directional predicted interpolation coding frame, also known as
(bi-directional interpolated prediction frame), it needs to refer to front and rear frame of video, utilizes two
The frame on side removes redundancy with the similitude of B frames.Inter prediction is mainly based upon higher similarity between consecutive frame, utilizes
The information of given frame rebuilds present frame, and commonly used approach is pre- to carry out interframe by way of motor-function evaluation
Survey.The method i.e. estimation method of the macro block most matched is found, this method is to be directed to current macro, first finds most phase in former frame
As macro block (i.e. match block), poor are asked to them, many null values are produced, carrying out the coding flow similar to frame in can with regard to that can save
The bit number of sight.
In mixed video coding framework, the frame of video of different frame type, its coding method is different, just because such as
This, its nature of distortion of the frame of video of different frame type is also different.Because I frames are relatively independent, therefore master the reason for its distortion
To be quantizing process and dct transform (discrete cosine transform) process, quantizing process is substantially that many numerical value are mapped into a small number of values
Process, relatively, de-quantization process recovers the process of many numerical value by a small number of values, and its recovery inevitably has number
According to error, therefore frame of video occurs some by quantifying the high-frequency noise that loss is caused.Producing the major way of P frames and B frames is
By way of inter prediction, the source of both frames prediction is from I frames, therefore the nature of distortion that I frames have in itself
Also P frames and B frames can be broadcast to；In addition, during motor-function evaluation, there is also the factor for causing distortion.Such as, in frame
Between the motion compensation link predicted, two neighboring piece of prediction block of present frame is possible to be not from the phase of same reference frame
The prediction of adjacent two blocks, it may be possible to from the prediction of two blocks in different frame.That is, that two of adjacent block reference
The edge of block does not just possess continuity in itself, and this also leads to certain blocking effect phenomenon, so as to cause distortion.
The general principle of Video coding is exactly the correlation that has using video sequence to compress redundancy, and realization is regarded
The high efficiency of transmission of frequency sequence.The correlation that wherein video sequence has is broadly divided into room and time correlation, spatially
Correlation has correlation between being mainly reflected in same frame of video neighbor pixel, and temporal correlation is mainly reflected in
Adjacent frame has higher similarity on time.Therefore, in mixed video coding framework, the generation side of each frame of video
Formula, is not the result of infra-frame prediction, is exactly the result of inter prediction.Its generation mechanism of different types of frame is also differed, meaning
Needs frame type being also contemplated for into the video encoding quality enhancing algorithm based on convolutional neural networks, to obtain more preferably
Effect.
Based on foregoing principle, it is rearmounted that the specific embodiment of the invention proposes the Video coding based on convolutional neural networks
Filtering method, this method includes the training step and post-filtering process step of convolutional neural networks model, as shown in Figure 1：
The training step includes S1 to S4：
S1, the quantization parameter of setting video compress are 20 to 51, carry out coding compression to original video, obtain compression and regard
Frequently；
S2, the extraction that frame is carried out to the compression video and the original video, obtain multiple frames pair, each frame pair
Include a compressed video frame and an original video frame；
S3, the frame for extracting step S2 are multiple groups to the different demarcation according to frame type and quantization parameter；
S4, convolutional neural networks framework and initialization network parameter are built, the group divided using step S3 is respectively to nerve
Network is trained, and obtains corresponding to different quantization parameters and multiple neural network models of frame type；
The post-filtering process step includes S5 and S6：
S5, the post-filtering link that multiple neural network models that step S4 is obtained are embedded in video encoder；
S6, step S1 and S2 are performed to pending original video obtain pending frame pair, and according to pending frame pair
Quantization parameter and the corresponding neural network model of frame type selecting are filtered processing.
The method of the present invention is filtered place firstly the need of training neutral net, then with the neural network model trained
Reason.Illustrate that the above-mentioned Video coding based on convolutional neural networks that the present invention is provided is rearmounted with a kind of specific embodiment below
Filtering method.
The original video for training is chosen, coding compression is carried out, quantization parameter QP during compression is set to 20 to 51 (i.e.
Continuous integer 20,21,22,23,24 ..., 50,51), obtain the compression video under different Q P.Preferably, can be using normal
JM10 and/or HM12 video encoding standards software carries out coding compression to original video.To original video and corresponding pressure
Contracting video carries out frame extraction, obtains many frames pair, and frame is to that can be expressed as " original video frame-compressed video frame ", i.e., each
Frame is to including an original video frame and a corresponding compressed video frame.By obtained all frames to according to frame type and quantization parameter
QP different demarcation is multiple groups, and specific partition process is, for example,：All frames are drawn to elder generation according to QP difference (32)
Point, then to the frame under each QP to being divided into I frames, P frames and B frames according to frame type, so as to obtain described multiple groups (in this example
In, obtain), the frame in each group is to identical frame type and quantization parameter QP.
The framework of neutral net to be trained includes convolutional layer and Relu layers, in a kind of specific embodiment, builds
The framework of neutral net refers to Fig. 2, including 3 convolutional layers (Convolution layer1, Convolution in figure
Layer2, Convolution layer3) and 2 Relu layers (ReLU layer1 and the ReLU layer2 in figure), convolutional layer
Alternately connected successively with Relu layers, for example, connecting convolutional layer 1 (Convolution layer1), convolution after input layer input
The output connection Relu layers 1 (ReLU layer1) of layer 1, the output connection convolutional layer 2 (Convolution of Relu layers 1
Layer2), convolutional layer 2 (Convolution layer2) output connection Relu layers 2 (ReLU layer2), Relu layers 2 it is defeated
Go out to connect convolutional layer 3 (Convolution layer3), the output connection output layer output of convolutional layer 3.It is to be appreciated that Fig. 2
What shown convolutional neural networks were merely exemplary, it is not construed as limiting the invention.To in the neutral net shown in Fig. 2,
The filtering core size (filtering core is convolution kernel) for setting convolutional layer 1,2,3 is respectively 9,1,5, neuron number is respectively 64,32,
1；Relu layers of neuron number with and the Relu layers of neuron number of last layer convolutional layer being connected be consistent.
Then, using foregoing obtained many framings to separately go training it is above-mentioned put up for example shown in Fig. 2
Neutral net.First, by the frame in each group to dividing training set, checking collection and test set, example according to resolution ratio and scene
Such as by QP be 20 and frame type be I frames all frames to be divided into training set, checking collection and test set, by QP be 20 and frame type
Collect and test set to being divided into training set, checking for all frames of P frames, wherein, the frame in test set is to resolution ratio and field
The characteristics of scape is various.Secondly, by all frames to according to certain pel spacing interception image block, obtaining " primitive frame image block-
The image block pair of condensed frame image block ", is used as the input of neutral net in training process.
Illustrate the process for training neutral net so that QP is the framing pair that 20, type is I frames as an example, for example, to the group
The frame of interior all frame centerings is that spacing carrys out interception image block by 28 pixels, such as the size of each image block is 33 × 33,
In the training process, all convolutional layers all avoid boundary effect (because using padding phases without using padding operations
When the pixel on border in imagineering, so as to expand the size of image, the information of boundary is just inaccurate), in order to
The penalty values between output image block and original picture block are calculated, original picture block is also cut into output image block by the present invention
It is in the same size.Then, all image blocks of interception are divided into multiple batches at random, the image block of a batch is sequentially input
Neutral net, it is then an iteration to complete a batch, and how many batch is to carry out how many times iteration, sharp in an iterative process
Back-propagating is carried out with stochastic gradient descent method, network parameter is updated, constantly loss function is minimized.Per the certain number of times of iteration
The network parameter trained is then corrected using checking collection, to prevent over-fitting.
Specific training process is as shown in figure 3, operation of the neuron to the image block of input of convolutional layer is expressed asWherein n represents the quantity of input picture block, wiRepresent i-th of convolution kernel, xiRepresent i-th of input picture
Block, b represents biasing coefficient, and * represents convolution operation, and image block is represented with picture element matrix；Relu layers of neuron is to image block
Operation be expressed as M=max (N, 0), wherein N and M represent the pixel value of input picture block and output image block respectively.
When parameter restrains, training is completed, it is the framing that 20, type is I frames to corresponding neutral net to obtain QP
Model, is embedded into encoder post-filtering link and is used as filtering.Similarly, QP is a framing that 20, type is P frames to also adopting
Neutral net is trained in aforementioned manners, obtains corresponding to the neural network model that QP is the framing pair that 20, type is P frames；
Frame under other QP values using same method to also training neutral net.
Assuming that the compression video employed in training process has two kinds of softwares of JM10 and HM12, then two kinds of Software Compressions are obtained
Compression video should separate and handled, separate training.So as to by above-mentioned training method, then obtain following
Multiple neural network models：
32 different I frame models that the lower correspondence QP of JM10 standard softwares coding is 20 to 51, JM10 standard softwares coding
32 different P frame models that lower correspondence QP is 20 to 51, the lower correspondence QP of JM10 standard softwares coding for 32 of 20 to 51 not
Same B frame models, 32 different I frame models that the lower correspondence QP of HM12 standard softwares coding is 20 to 51, HM12 standard softwares
32 different P frame models that the lower correspondence QP of coding is 20 to 51, the lower correspondence QP of HM12 standard softwares coding is the 32 of 20 to 51
Individual different B frame models.
In actual use, coding compression and frame are carried out to extracting it to the pending encoded device of original video
Afterwards, obtained frame is for example judged to learn that the QP of currently pending frame pair is to carrying out the judgement of quantization parameter and frame type
23rd, type is B frames, then it is that the image block that 23, type is B frames trains obtained neural network model come after carrying out that selection, which belongs to QP,
Put filtering process.In a preferred embodiment, the encoding software of use can be also contemplated for into, i.e., in training always according to coding
The difference of software is trained (as described by the preceding paragraph) respectively, so that it is which kind of that first resolution is further contemplated in actual use
Encoding software compressed encoding is obtained, and considers further that quantization parameter and frame type to select corresponding neural network model to be filtered
Processing.
Above content is to combine specific preferred embodiment further description made for the present invention, it is impossible to assert
The specific implementation of the present invention is confined to these explanations.For those skilled in the art, do not taking off
On the premise of from present inventive concept, some equivalent substitutes or obvious modification can also be made, and performance or purposes are identical, all should
When being considered as belonging to protection scope of the present invention.
Claims (8)
1. a kind of Video coding post-filter method based on convolutional neural networks, includes the training step of convolutional neural networks model
Rapid and post-filtering process step, wherein：
The training step includes S1 to S4：
S1, the quantization parameter of setting video compress are 20 to 51, and coding compression is carried out to original video, obtain compressing video；
S2, the extraction that frame is carried out to the compression video and the original video, obtain multiple frames pair, each frame to comprising
One compressed video frame and an original video frame；
S3, the frame for extracting step S2 are multiple groups to the different demarcation according to frame type and quantization parameter；
S4, convolutional neural networks framework and initialization network parameter are built, the group divided using step S3 is respectively to neutral net
It is trained, obtains corresponding to different quantization parameters and multiple neural network models of frame type；
The post-filtering process step includes S5 and S6：
S5, the post-filtering link that multiple neural network models that step S4 is obtained are embedded in video encoder；
S6, step S1 and S2 are performed to pending original video obtain pending frame pair, and according to the quantization of pending frame pair
Parameter and the corresponding neural network model of frame type selecting are filtered processing.
2. the Video coding post-filter method as claimed in claim 1 based on convolutional neural networks, it is characterised in that：Step
By frame to being divided into I frames, P frames and B frames according to frame type in S3.
3. the Video coding post-filter method as claimed in claim 1 based on convolutional neural networks, it is characterised in that：Step
S3 is also included the frame in each group to dividing training set, checking collection and test set according to resolution ratio and scene；It is wherein described
Checking collection is used for the parameter trained in the training process per iteration pre-determined number post-equalization, to prevent over-fitting.
4. the Video coding post-filter method as claimed in claim 3 based on convolutional neural networks, it is characterised in that：It is described
Frame pair including a variety of resolution ratio and several scenes in test set.
5. the Video coding post-filter method as claimed in claim 1 based on convolutional neural networks, it is characterised in that：Step
When S4 carries out neural metwork training, by frame to according to predetermined pel spacing interception image block, obtaining image block to as training
During neutral net input；Wherein, each image block is to including a primitive frame image block and a condensed frame image block.
6. the Video coding post-filter method as claimed in claim 5 based on convolutional neural networks, it is characterised in that：Step
The framework for the convolutional neural networks built in S4 includes convolutional layer and Relu layers；Wherein：
Operation of the neuron of convolutional layer to the image block of input is expressed asWherein n represents input picture block
Quantity, wiRepresent i-th of convolution kernel, xiI-th of input picture block is represented, b represents to bias coefficient, and * represents convolution operation, schemes
As block is represented with picture element matrix；
Operation of Relu layers of the neuron to image block is expressed as M=max (N, 0), and wherein N and M represent input picture block respectively
With the pixel value of output image block.
7. the Video coding post-filter method as claimed in claim 6 based on convolutional neural networks, it is characterised in that：Convolution
Layer and Relu layers of alternately connection successively.
8. the Video coding post-filter method as claimed in claim 5 based on convolutional neural networks, it is characterised in that：Nerve
During network training, the size of the original picture block of input is cut in the same size with output image block.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710439132.8A CN107197260B (en) | 2017-06-12 | 2017-06-12 | Video coding post-filter method based on convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710439132.8A CN107197260B (en) | 2017-06-12 | 2017-06-12 | Video coding post-filter method based on convolutional neural networks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107197260A true CN107197260A (en) | 2017-09-22 |
| CN107197260B CN107197260B (en) | 2019-09-13 |
Family
ID=59876636
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710439132.8A Active CN107197260B (en) | 2017-06-12 | 2017-06-12 | Video coding post-filter method based on convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107197260B (en) |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107743235A (en) * | 2017-10-27 | 2018-02-27 | 厦门美图之家科技有限公司 | Image processing method, device and electronic equipment |
| CN108134932A (en) * | 2018-01-11 | 2018-06-08 | 上海交通大学 | Filter achieving method and system in coding and decoding video loop based on convolutional neural networks |
| CN108174225A (en) * | 2018-01-11 | 2018-06-15 | 上海交通大学 | Filter achieving method and system in coding and decoding video loop based on confrontation generation network |
| CN108289224A (en) * | 2017-12-12 | 2018-07-17 | 北京大学 | A kind of video frame prediction technique, device and neural network is compensated automatically |
| CN108307193A (en) * | 2018-02-08 | 2018-07-20 | 北京航空航天大学 | A kind of the multiframe quality enhancement method and device of lossy compression video |
| CN109451308A (en) * | 2018-11-29 | 2019-03-08 | 北京市商汤科技开发有限公司 | Video compression method and device, electronic equipment and storage medium |
| WO2019141193A1 (en) * | 2018-01-19 | 2019-07-25 | 杭州海康威视数字技术股份有限公司 | Method and apparatus for processing video frame data |
| WO2019141255A1 (en) * | 2018-01-18 | 2019-07-25 | 杭州海康威视数字技术股份有限公司 | Image filtering method and device |
| WO2019141258A1 (en) * | 2018-01-18 | 2019-07-25 | 杭州海康威视数字技术股份有限公司 | Video encoding method, video decoding method, device, and system |
| CN110351568A (en) * | 2019-06-13 | 2019-10-18 | 天津大学 | A kind of filtering video loop device based on depth convolutional network |
| WO2019205871A1 (en) * | 2018-04-25 | 2019-10-31 | 杭州海康威视数字技术股份有限公司 | Image decoding and encoding methods and apparatuses, and device thereof |
| CN110740319A (en) * | 2019-10-30 | 2020-01-31 | 腾讯科技（深圳）有限公司 | Video encoding and decoding method and device, electronic equipment and storage medium |
| CN110971915A (en) * | 2018-09-28 | 2020-04-07 | 杭州海康威视数字技术股份有限公司 | Filtering method and device |
| CN111083482A (en) * | 2019-12-31 | 2020-04-28 | 合肥图鸭信息科技有限公司 | Video compression network training method and device and terminal equipment |
| CN111107357A (en) * | 2018-10-25 | 2020-05-05 | 杭州海康威视数字技术股份有限公司 | Image processing method, device and system |
| CN111212288A (en) * | 2020-01-09 | 2020-05-29 | 广州虎牙科技有限公司 | Video data encoding and decoding method and device, computer equipment and storage medium |
| CN111556337A (en) * | 2020-05-15 | 2020-08-18 | 腾讯科技（深圳）有限公司 | Media content implantation method, model training method and related device |
| CN111699686A (en) * | 2018-01-26 | 2020-09-22 | 联发科技股份有限公司 | Method and device for packet neural network for video coding and decoding |
| CN111726613A (en) * | 2020-06-30 | 2020-09-29 | 福州大学 | Video coding optimization method based on just noticeable difference |
| CN111742553A (en) * | 2017-12-14 | 2020-10-02 | 交互数字Vc控股公司 | Deep learning based image partitioning for video compression |
| CN111937392A (en) * | 2018-04-17 | 2020-11-13 | 联发科技股份有限公司 | Neural network method and device for video coding and decoding |
| CN112019843A (en) * | 2019-05-30 | 2020-12-01 | 富士通株式会社 | Encoding and decoding program, encoding and decoding device, encoding and decoding method |
| CN112470472A (en) * | 2018-06-11 | 2021-03-09 | 无锡安科迪智能技术有限公司 | Blind compression sampling method and device and imaging system |
| WO2021064291A1 (en) * | 2019-10-02 | 2021-04-08 | Nokia Technologies Oy | Guiding decoder-side optimization of neural network filter |
| WO2021109846A1 (en) * | 2019-12-06 | 2021-06-10 | 华为技术有限公司 | Bit stream data processing method and apparatus |
| CN113259671A (en) * | 2020-02-10 | 2021-08-13 | 腾讯科技（深圳）有限公司 | Loop filtering method, device and equipment in video coding and decoding and storage medium |
| CN113542867A (en) * | 2020-04-20 | 2021-10-22 | 声音猎手公司 | Content filtering in a media playback device |
| CN114125449A (en) * | 2021-10-26 | 2022-03-01 | 阿里巴巴新加坡控股有限公司 | Video processing method, system and computer readable medium based on neural network |
| CN115460463A (en) * | 2018-01-04 | 2022-12-09 | 三星电子株式会社 | Video playing device and control method thereof |
| WO2023116173A1 (en) * | 2021-12-21 | 2023-06-29 | 腾讯科技（深圳）有限公司 | Data processing method and apparatus, and computer device and storage medium |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114501031B (en) * | 2020-11-13 | 2023-06-02 | 华为技术有限公司 | Compression coding and decompression method and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6075884A (en) * | 1996-03-29 | 2000-06-13 | Sarnoff Corporation | Method and apparatus for training a neural network to learn and use fidelity metric as a control mechanism |
| US7822698B1 (en) * | 2007-03-23 | 2010-10-26 | Hrl Laboratories, Llc | Spike domain and pulse domain non-linear processors |
| CN105160678A (en) * | 2015-09-02 | 2015-12-16 | 山东大学 | Convolutional-neural-network-based reference-free three-dimensional image quality evaluation method |
| CN105894046A (en) * | 2016-06-16 | 2016-08-24 | 北京市商汤科技开发有限公司 | Convolutional neural network training and image processing method and system and computer equipment |
| CN106485661A (en) * | 2016-11-15 | 2017-03-08 | 杭州当虹科技有限公司 | A kind of high-quality image magnification method |
| CN106682697A (en) * | 2016-12-29 | 2017-05-17 | 华中科技大学 | End-to-end object detection method based on convolutional neural network |
-
2017
- 2017-06-12 CN CN201710439132.8A patent/CN107197260B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6075884A (en) * | 1996-03-29 | 2000-06-13 | Sarnoff Corporation | Method and apparatus for training a neural network to learn and use fidelity metric as a control mechanism |
| US7822698B1 (en) * | 2007-03-23 | 2010-10-26 | Hrl Laboratories, Llc | Spike domain and pulse domain non-linear processors |
| CN105160678A (en) * | 2015-09-02 | 2015-12-16 | 山东大学 | Convolutional-neural-network-based reference-free three-dimensional image quality evaluation method |
| CN105894046A (en) * | 2016-06-16 | 2016-08-24 | 北京市商汤科技开发有限公司 | Convolutional neural network training and image processing method and system and computer equipment |
| CN106485661A (en) * | 2016-11-15 | 2017-03-08 | 杭州当虹科技有限公司 | A kind of high-quality image magnification method |
| CN106682697A (en) * | 2016-12-29 | 2017-05-17 | 华中科技大学 | End-to-end object detection method based on convolutional neural network |
Cited By (47)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107743235A (en) * | 2017-10-27 | 2018-02-27 | 厦门美图之家科技有限公司 | Image processing method, device and electronic equipment |
| CN108289224B (en) * | 2017-12-12 | 2019-10-29 | 北京大学 | A kind of video frame prediction technique, device and neural network is compensated automatically |
| CN108289224A (en) * | 2017-12-12 | 2018-07-17 | 北京大学 | A kind of video frame prediction technique, device and neural network is compensated automatically |
| CN111742553A (en) * | 2017-12-14 | 2020-10-02 | 交互数字Vc控股公司 | Deep learning based image partitioning for video compression |
| CN115460463A (en) * | 2018-01-04 | 2022-12-09 | 三星电子株式会社 | Video playing device and control method thereof |
| CN108174225B (en) * | 2018-01-11 | 2021-03-26 | 上海交通大学 | Video coding and decoding in-loop filtering implementation method and system based on countermeasure generation network |
| CN108134932B (en) * | 2018-01-11 | 2021-03-30 | 上海交通大学 | Method and system for realizing video coding and decoding loop internal filtering based on convolutional neural network |
| CN108174225A (en) * | 2018-01-11 | 2018-06-15 | 上海交通大学 | Filter achieving method and system in coding and decoding video loop based on confrontation generation network |
| CN108134932A (en) * | 2018-01-11 | 2018-06-08 | 上海交通大学 | Filter achieving method and system in coding and decoding video loop based on convolutional neural networks |
| WO2019141255A1 (en) * | 2018-01-18 | 2019-07-25 | 杭州海康威视数字技术股份有限公司 | Image filtering method and device |
| WO2019141258A1 (en) * | 2018-01-18 | 2019-07-25 | 杭州海康威视数字技术股份有限公司 | Video encoding method, video decoding method, device, and system |
| WO2019141193A1 (en) * | 2018-01-19 | 2019-07-25 | 杭州海康威视数字技术股份有限公司 | Method and apparatus for processing video frame data |
| CN110062246A (en) * | 2018-01-19 | 2019-07-26 | 杭州海康威视数字技术股份有限公司 | The method and apparatus that video requency frame data is handled |
| CN111699686A (en) * | 2018-01-26 | 2020-09-22 | 联发科技股份有限公司 | Method and device for packet neural network for video coding and decoding |
| WO2019154152A1 (en) * | 2018-02-08 | 2019-08-15 | 北京航空航天大学 | Multi-frame quality enhancement method and device for lossy compressed video |
| CN108307193A (en) * | 2018-02-08 | 2018-07-20 | 北京航空航天大学 | A kind of the multiframe quality enhancement method and device of lossy compression video |
| CN108307193B (en) * | 2018-02-08 | 2018-12-18 | 北京航空航天大学 | A kind of the multiframe quality enhancement method and device of lossy compression video |
| US10965959B2 (en) | 2018-02-08 | 2021-03-30 | Beihang University | Multi-frame quality enhancement for compressed video |
| CN111937392A (en) * | 2018-04-17 | 2020-11-13 | 联发科技股份有限公司 | Neural network method and device for video coding and decoding |
| CN110401836B (en) * | 2018-04-25 | 2022-04-26 | 杭州海康威视数字技术股份有限公司 | Image decoding and encoding method, device and equipment |
| WO2019205871A1 (en) * | 2018-04-25 | 2019-10-31 | 杭州海康威视数字技术股份有限公司 | Image decoding and encoding methods and apparatuses, and device thereof |
| CN110401836A (en) * | 2018-04-25 | 2019-11-01 | 杭州海康威视数字技术股份有限公司 | A kind of image decoding, coding method, device and its equipment |
| CN112470472A (en) * | 2018-06-11 | 2021-03-09 | 无锡安科迪智能技术有限公司 | Blind compression sampling method and device and imaging system |
| CN110971915A (en) * | 2018-09-28 | 2020-04-07 | 杭州海康威视数字技术股份有限公司 | Filtering method and device |
| CN111107357B (en) * | 2018-10-25 | 2022-05-31 | 杭州海康威视数字技术股份有限公司 | Image processing method, device, system and storage medium |
| CN111107357A (en) * | 2018-10-25 | 2020-05-05 | 杭州海康威视数字技术股份有限公司 | Image processing method, device and system |
| CN109451308A (en) * | 2018-11-29 | 2019-03-08 | 北京市商汤科技开发有限公司 | Video compression method and device, electronic equipment 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 |
| CN109451308B (en) * | 2018-11-29 | 2021-03-09 | 北京市商汤科技开发有限公司 | Video compression processing method and device, electronic equipment and storage medium |
| CN112019843A (en) * | 2019-05-30 | 2020-12-01 | 富士通株式会社 | Encoding and decoding program, encoding and decoding device, encoding and decoding method |
| CN112019843B (en) * | 2019-05-30 | 2022-07-15 | 富士通株式会社 | Encoding and decoding method, apparatus, readable storage medium |
| CN110351568A (en) * | 2019-06-13 | 2019-10-18 | 天津大学 | A kind of filtering video loop device based on depth convolutional network |
| WO2021064291A1 (en) * | 2019-10-02 | 2021-04-08 | Nokia Technologies Oy | Guiding decoder-side optimization of neural network filter |
| US11341688B2 (en) | 2019-10-02 | 2022-05-24 | Nokia Technologies Oy | Guiding decoder-side optimization of neural network filter |
| CN110740319B (en) * | 2019-10-30 | 2024-04-05 | 腾讯科技（深圳）有限公司 | Video encoding and decoding method and device, electronic equipment and storage medium |
| CN110740319A (en) * | 2019-10-30 | 2020-01-31 | 腾讯科技（深圳）有限公司 | Video encoding and decoding method and device, electronic equipment and storage medium |
| WO2021109846A1 (en) * | 2019-12-06 | 2021-06-10 | 华为技术有限公司 | Bit stream data processing method and apparatus |
| CN111083482A (en) * | 2019-12-31 | 2020-04-28 | 合肥图鸭信息科技有限公司 | Video compression network training method and device and terminal equipment |
| CN111212288B (en) * | 2020-01-09 | 2022-10-04 | 广州虎牙科技有限公司 | Video data encoding and decoding method and device, computer equipment and storage medium |
| CN111212288A (en) * | 2020-01-09 | 2020-05-29 | 广州虎牙科技有限公司 | Video data encoding and decoding method and device, computer equipment and storage medium |
| CN113259671A (en) * | 2020-02-10 | 2021-08-13 | 腾讯科技（深圳）有限公司 | Loop filtering method, device and equipment in video coding and decoding and storage medium |
| CN113542867A (en) * | 2020-04-20 | 2021-10-22 | 声音猎手公司 | Content filtering in a media playback device |
| CN111556337A (en) * | 2020-05-15 | 2020-08-18 | 腾讯科技（深圳）有限公司 | Media content implantation method, model training method and related device |
| CN111726613A (en) * | 2020-06-30 | 2020-09-29 | 福州大学 | Video coding optimization method based on just noticeable difference |
| CN111726613B (en) * | 2020-06-30 | 2021-07-27 | 福州大学 | Video coding optimization method based on just noticeable difference |
| CN114125449A (en) * | 2021-10-26 | 2022-03-01 | 阿里巴巴新加坡控股有限公司 | Video processing method, system and computer readable medium based on neural network |
| WO2023116173A1 (en) * | 2021-12-21 | 2023-06-29 | 腾讯科技（深圳）有限公司 | Data processing method and apparatus, and computer device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107197260B (en) | 2019-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107197260B (en) | Video coding post-filter method based on convolutional neural networks | |
| CN106960416A (en) | A kind of video satellite compression image super-resolution method of content complexity self adaptation | |
| CN106716997A (en) | Video encoding method using in-loop filter parameter prediction and apparatus therefor, and video decoding method and apparatus therefor | |
| CN110351568A (en) | A kind of filtering video loop device based on depth convolutional network | |
| CN104093021B (en) | Monitoring video compression method | |
| CN103782598A (en) | Fast encoding method for lossless coding | |
| CN108184118A (en) | Cloud desktop contents encode and coding/decoding method and device, system | |
| CN108900848A (en) | A kind of video quality Enhancement Method based on adaptive separable convolution | |
| CN113766249B (en) | Loop filtering method, device, equipment and storage medium in video coding and decoding | |
| CN111031315B (en) | Compressed video quality enhancement method based on attention mechanism and time dependence | |
| CN101272489B (en) | Encoding and decoding device and method for video image quality enhancement | |
| CN105791877A (en) | Adaptive loop filter method in video coding and decoding | |
| CN111247797A (en) | Method and apparatus for image encoding and decoding | |
| CN103347185A (en) | Unmanned aerial vehicle scouting image synthesis compressed encoding method based on selective block transformation | |
| CN109903351A (en) | The method for compressing image combined based on convolutional neural networks and traditional code | |
| Li et al. | Multi-scale grouped dense network for vvc intra coding | |
| CN112188217B (en) | JPEG compressed image decompression effect removing method combining DCT domain and pixel domain learning | |
| CN113055674B (en) | Compressed video quality enhancement method based on two-stage multi-frame cooperation | |
| KR100697516B1 (en) | Moving picture coding method based on 3D wavelet transformation | |
| CN110677644A (en) | Video coding and decoding method and video coding intra-frame predictor | |
| CN103747257B (en) | A kind of method of video data high efficient coding | |
| CN112001854A (en) | Method for repairing coded image and related system and device | |
| KR101998036B1 (en) | Artifact reduction method and apparatus | |
| CN103313064B (en) | Based on the time domain error hidden method of inter-frame mode and movement repair | |
| CN115002482B (en) | End-to-end video compression method and system using structural preserving motion estimation |
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 |