CN108012157A - Construction method for the convolutional neural networks of Video coding fractional pixel interpolation - Google Patents
Construction method for the convolutional neural networks of Video coding fractional pixel interpolation Download PDFInfo
- Publication number
- CN108012157A CN108012157A CN201711207766.7A CN201711207766A CN108012157A CN 108012157 A CN108012157 A CN 108012157A CN 201711207766 A CN201711207766 A CN 201711207766A CN 108012157 A CN108012157 A CN 108012157A
- Authority
- CN
- China
- Prior art keywords
- convolutional neural
- neural networks
- fractional pixel
- video coding
- pixel interpolation
- 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
- H04N19/80—Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- 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
- H04N19/132—Sampling, masking or truncation of coding units, e.g. adaptive resampling, frame skipping, frame interpolation or high-frequency transform coefficient masking
-
- 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
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/587—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal sub-sampling or interpolation, e.g. decimation or subsequent interpolation of pictures in a video sequence
-
- 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
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/625—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using discrete cosine transform [DCT]
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Discrete Mathematics (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Image Processing (AREA)
Abstract
The present invention provides a kind of construction method of convolutional neural networks for Video coding fractional pixel interpolation, including:Different content, the image of resolution ratio are collected, is formed comprising different type, the original training data collection of the data of encoder complexity;Pretreatment operation is carried out to original training data collection, obtains the training data for meeting Video coding inter prediction fractional pixel interpolation characteristic;Depth convolutional neural networks are built, obtain the convolutional neural networks structure suitable for Video coding inter prediction fractional pixel interpolation;The convolutional neural networks that the data input obtained using pretreatment is put up, while the convolutional neural networks that original training data collection is built as corresponding true value, training.This invention ensures that convolutional neural networks can be trained smoothly, and meeting Video coding fractional pixel interpolation property requirements using the fraction pixel that trained convolutional neural networks interpolation obtains, the lifting of video coding efficiency can be realized by carrying out fractional pixel interpolation using the present invention.
Description
Technical field
The present invention relates to a kind of method of technical field of image processing, is specifically that one kind is suitable for Video coding inter prediction
The convolutional neural networks method of fractional pixel interpolation.
Background technology
Inter prediction is a key technology in video encoding standard, using between frame and frame video content it is similar
Property, the redundancy of video in time can be effectively removed, so as to improve coding compression efficiency.Simultaneously as digitizing
Discrete sampling operation in journey, real object of which movement are not necessarily what is carried out according to sampling grid.In order to further improve thing
The accuracy of body motion prediction, the movement of object is all in units of fraction pixel in video encoding standard.Sampling grid
The upper pixel value positioned at fractional pixel position is not necessary being, and in the application, the pixel value of these fractional pixel positions needs
The pixel value interpolation of the integer position of necessary being is utilized to obtain.
However, it is based on some a priori assumptions that the interpolation filter that fraction pixel is used is generated in Video coding at present
On the basis of, artificially design.The parameter of these interpolation filters is fixed, enriching constantly and regarding with video content
Frequency division resolution is continuously increased, and the wave filter of this preset parameter can not be all applicable in.
Deep learning is mass data to be fitted by the neutral net of design so as to obtain one kind of universal applicable models
Method.Method based on deep learning not only for example achieves great in some semantic class problems in target following, pedestrian detection
Break through, effect has also been obviously improved in the Pixel-level problem such as image super-resolution.
Inter prediction fractional pixel interpolation has certain similitude with image super-resolution, i.e., both by necessary being
Small figure by the big figure of certain multiplying power generation.But image super-resolution is to generate whole high-resolution using low-resolution image
Big figure, and inter prediction fractional pixel interpolation is then to generate remaining fractional position picture according to the integer position pixel of necessary being
Element is, it is necessary to ensure that integer position pixel does not change.In addition, for inter prediction fractional pixel interpolation, positioned at the picture of fractional position
Element is not necessary being, therefore, in the training process of convolutional neural networks, may be referred to, leads without real true value
Training is caused to be normally carried out.
The content of the invention
The present invention is in view of the foregoing defects the prior art has, there is provided one kind is suitable for Video coding inter prediction fraction picture
The construction method of the convolutional neural networks of plain interpolation, this method utilize the volume that superperformance is obtained in image super-resolution problem
The advantages of product neutral net, while the characteristics of consider Video coding inter prediction fractional pixel interpolation, devise suitable for video
Encode the convolutional neural networks of inter prediction fractional pixel interpolation and make the pretreatment operation that is smoothed out of training, so can be with
The objective quality of Video coding reconstruction frames is improved, realizes the lifting of code efficiency.
To achieve the above object, the structure of the convolutional neural networks of the present invention for Video coding fractional pixel interpolation
Construction method includes:
Different content, the image of different resolution are collected, is formed comprising different type, the data of different coding complexity
Original training data collection;
Pretreatment operation is carried out to the original training data collection being collected into, obtains meeting Video coding inter prediction fraction picture
The training data of plain interpolation characteristic, input data of the data as training convolutional neural networks;
Depth convolutional neural networks are built, Video coding fractional pixel interpolation characteristic is considered, obtains being suitable for Video coding
The convolutional neural networks structure of inter prediction fractional pixel interpolation;
The convolutional neural networks put up of data input obtained using pretreatment, while by the original training data collection
The convolutional neural networks built as corresponding true value, training, obtain being suitable for Video coding inter prediction fractional pixel interpolation
Convolutional neural networks model.
Preferably, the pretreatment operation, process are as follows:
A) image that the fractional pixel position of interpolation generation concentrates original training data as needed carries out corresponding multiplying power
Down-sampled operation, obtain the low resolution training data for step b);
B) volume is compressed to low resolution training data according to the configuration in video encoding standard to still image coding
Code, obtains the low resolution coding and rebuilding image for step c);
C) up-sampling for carrying out corresponding to multiplying power to low resolution coding and rebuilding image in step a) operates, and returns to original graph
As size, the input data of training convolutional neural networks is obtained.
It is highly preferred that it is described c) in, the up-sampling of low resolution coding and rebuilding image is operated, is ensured high after up-sampling
The pixel value of image in different resolution integer pixel positions is consistent with the low resolution coding and rebuilding figure before up-sampling.
Preferably, it is described to build depth convolutional neural networks, wherein the depth convolution god network built includes 20 weights
Layer and 1 weight masking layer;For weight masking layer, WIFor the weight of integer pixel positions, WHFor fractional pixel position
Weight, all fractional pixel positions share a weight.
It is highly preferred that the Video coding inter prediction fractional pixel interpolation, wherein integer pixel positions pixel value is constant,
Only generate fractional pixel position.
Compared with prior art, the beneficial effects of the invention are as follows:
The present invention is extracted beyond the great ability of feature using depth convolutional neural networks from mass data, it is also contemplated that
Video coding distinctive data characteristic and Video coding inter prediction fractional pixel interpolation are only compared to image super-resolution
The characteristics of having, redesigned depth convolutional neural networks, while devises supporting pretreatment operation, ensures convolutional Neural net
The training of network can be smoothed out, so that the convolutional neural networks model suitable for Video coding fractional pixel interpolation has been obtained,
The objective quality that compressed encoding rebuilds video is improved, improves video coding efficiency.
Brief description of the drawings
Upon reading the detailed description of non-limiting embodiments with reference to the following drawings, further feature of the invention,
Objects and advantages will become more apparent upon:
Fig. 1 is the method flow diagram of one embodiment of the invention;
Fig. 2 is the convolutional neural networks structure diagram of one embodiment of the invention;
Fig. 3 is one embodiment of the invention integer pixel positions, half fractional pixel position, a quarter fraction pixel
Position view.
Embodiment
With reference to specific embodiment, the present invention is described in detail.Following embodiments will be helpful to the technology of this area
Personnel further understand the present invention, but the invention is not limited in any way.It should be pointed out that the ordinary skill to this area
For personnel, without departing from the inventive concept of the premise, various modifications and improvements can be made.These belong to the present invention
Protection domain.
The present invention provides a kind of construction method of convolutional neural networks for Video coding fractional pixel interpolation, such as Fig. 1
Shown, its mentality of designing is:
Different content, the image of different resolution are collected, is obtained comprising different type, the data of different coding complexity
Training dataset;
The training dataset being collected into is pre-processed, obtains the input data of training convolutional neural networks.Pretreatment
Operation specifically includes:
A) image that the fractional pixel position of interpolation generation concentrates original training data as needed carries out corresponding multiplying power
Down-sampled operation, obtain the low resolution training data for step b);
B) volume is compressed to low resolution training data according to the configuration in video encoding standard to still image coding
Code, obtains the low resolution coding and rebuilding image for step c);
C) up-sampling for carrying out corresponding to multiplying power to low resolution coding and rebuilding image in step a) operates, and returns to original graph
As size, the input data of training convolutional neural networks is obtained.
The depth convolutional neural networks suitable for Video coding inter prediction fractional pixel interpolation are built, pretreatment will be passed through
Operate the input of obtained image as network, while original training data concentrated into corresponding image as corresponding true value,
Training parameter, training convolutional neural networks are set;
The convolutional neural networks model obtained using training carries out fractional pixel interpolation operation, and realization is based on convolutional Neural net
The Video coding inter prediction fractional pixel interpolation of network.
The b of the pre-treatment step), according to the configuration in video encoding standard for still image compression coding, to drop
Low-resolution image after sampling is compressed coding, makes the reconstructed value of low-resolution image become special comprising video data encoder
The image of property.
The c of the pre-treatment step), operated for the up-sampling of low resolution reconstruction image after compressed encoding, it is necessary to protect
The pixel value of the whole location of pixels of high-definition picture is consistent with low-resolution image before up-sampling after card up-sampling, only generates
The pixel value of fractional pixel position.
The present invention considers consolidating for Video coding fractional pixel interpolation on the basis of image super-resolution convolutional neural networks
There is characteristic i.e. integer position pixel constant, only generate fractional position pixel, redesign convolutional neural networks, meanwhile, with closing
Pretreatment operation is stated, ensure that convolutional neural networks can be trained smoothly, and uses trained convolutional neural networks interpolation
Obtained fraction pixel meets Video coding fractional pixel interpolation property requirements so that carries out fractional pixel interpolation using the present invention
It can realize the lifting of video coding efficiency.In addition, the convolutional neural networks obtained using the present invention, can be in once-through operation
The pixel value of all fractional pixel positions is generated at the same time.
Newest video encoding standard is applied the invention to below --- in high-performance video coding (HEVC), introduce suitable
For the convolutional neural networks construction method of HEVC inter prediction half picture element interpolations, mainly to data prediction, volume
Product neural network structure the specific implementation details such as is built and is described in detail.Certainly, the present invention can also be applied to other compile
Code standard.
1. process of data preprocessing
For in process of data preprocessing to the compressed encoding step of low-resolution image, using (AI) in the full frame of HEVC
Configuration encodes down-sampled obtained low resolution image.
For in preprocessing process to it is low resolution compressed encoding reconstruction image upsampling process, using based on discrete cosine
The interpolation filter of conversion.For half location of pixels, the interpolation filter based on discrete cosine transform is 8 tap filterings
Device, tap coefficient are as shown in table 1.
Interpolation filter tap coefficient of the table 1 based on discrete cosine transform
Index i | -3 | -2 | -1 | 0 | 1 | 2 | 3 | 4 |
Hfilter[i] | -1 | 4 | -11 | 40 | 40 | -11 | 4 | -1 |
The process of the half location of pixels pixel in Fig. 3 is produced using the interpolation filter based on discrete cosine transform
It is as follows:
Wherein, b0,j,hi,0,j0,0, the pixel value of expression half location of pixels, Ai,jRepresent whole location of pixels pixel
Value, hfilter [i] represent the tap coefficient of the interpolation filter based on discrete cosine transform, and B represents locating depth for pixel value.
2. convolutional neural networks structure is built
The present invention is using J Kim etc. in IEEE Conference on Computer Vision and in 2016
The Accurate delivered in Pattern Recognition (IEEE international computers vision and pattern-recognition meeting) meeting
Image Super-Resolution Using Very Deep Convolutional Networks are basic framework, in original
Weight masking layer, W are added in beginning frameIFor the weighted value of integer position pixel value, WHFor half location of pixels pixel value
Weighted value.
As shown in Fig. 2, the convolutional neural networks structure that the present embodiment is built includes 20 convolutional layers, 1 weight masking layer.
For convolutional layer, in addition to first convolutional layer and last convolutional layer, each convolutional layer includes 64 different filtering
Device, the size of each wave filter is 3 × 3 × 64.For first convolutional layer, the wave filter that 64 sizes are 3 × 3 × 1 is included.
For last convolutional layer, the wave filter that 1 size is 3 × 3 × 64 is included.For weight masking layer, integer pixel positions
Different weights, wherein W are used from fractional pixel positionIFor integer pixel positions weights, WHWeighed for half location of pixels
Value.Convolutional neural networks input in the present embodiment is the height of the target size obtained by low-resolution image after pretreatment
Image in different resolution.What the convolutional neural networks in the present embodiment were predicted is that the high-definition picture of final output and starting input
By the residual image between pretreatment image, it is defined as follows:
R=YH-XILR (4)
Wherein YHRepresent the high-definition picture of final output, XILRImage after the pretreatment of expression starting input.
By the way that the residual image that convolutional neural networks are predicted is added with input pretreatment image, final output is obtained
High-definition picture.
3. training convolutional neural networks
The training process of convolutional neural networks is using Euclidean distance as loss function:
Wherein θ represents that convolutional neural networks need the parameter set learnt,Represent training image,Represent original
Training data concentrates corresponding true value image, F (Xi;θ) represent the high-definition picture of final output.By rolling up in this present embodiment
Product neural network prediction is residual image, the F (X in formula (5)i;It should θ) be expressed as:
Wherein,Represent the image by pretreatment of starting input.
Training obtains the convolutional neural networks model suitable for Video coding inter prediction fractional pixel interpolation above.
4. implementation result
The convolutional neural networks model that the present embodiment is trained is applied in HEVC coding frameworks, use is improved
Encoder encodes cycle tests with standard HEVC encoders.Cycle tests is as shown in table 2, and all cycle tests are all 4:
2:0 yuv format, it is 8 to represent locating depth.
2 cycle tests details of table
For the HEVC encoders used in the present embodiment for HM-16.7, coding is configured to low latency P frames (LDP) universal test
Configuration, the quantization parameter (QP) for encoding use is respectively that 22,27,32,37. the present embodiment are based on just for luminance Y component use
The fractional pixel interpolation method of convolutional neural networks, remaining chromatic component is still using standard interpolation filter generation fraction picture
Element.
Under above-mentioned implementation condition, the encoded test result shown in table 3 has been obtained.The performance indicator that table 3 uses is BD-
Rate indexs, expression is compared with standard HEVC encoders, in the case of identical Y-PSNR (PSNR), uses this
The convolutional neural networks that embodiment is trained carry out the percentage that inter prediction half fractional pixel interpolation code check is saved.
As shown in table 3, under above-mentioned implementation condition, the average BD-Rate of tri- components of Y, U, V is respectively -0.9%, -0.1%, -
0.1%.Especially, the gain of sequence B asketballPass is most notable, the three-component gain of Y, U, V can reach -2.4%, -
0.1%th, -1.6%.From table 3 it can be seen that compared to standard HEVC encoders, instructed using luminance Y component is directed in the present embodiment
The method that the convolutional neural networks got carry out luminance component half picture element interpolation has obvious code efficiency to be lifted.
Further, since encoder has used the technology based on luminance component prediction chromatic component, with carrying for luminance component reconstruction quality
Rise, remaining chromatic component can also obtain certain coding efficiency lifting.
3 cycle tests coding efficiency (BD-Rate) of table
To further illustrate that the convolutional neural networks of present invention structure is more suitable for the fraction in Video coding inter prediction
Picture element interpolation is direct shown in table 4 using two points of the convolutional neural networks trained for image super-resolution problem progress
One of test result of the fractional pixel interpolation compared with using standard HEVC encoders.From table 4, it can be seen that directly use image
The convolutional neural networks of super-resolution, which carry out fractional pixel interpolation, obvious loss of coding performance.
Table 4 uses image super-resolution convolutional neural networks encoded test result (BD-Rate)
To sum up, the present invention devises special convolutional neural networks for Video coding inter prediction fractional pixel interpolation,
Meanwhile the present invention devises supporting process of data preprocessing so that the training of convolutional neural networks can be smoothed out, and
The fraction pixel generated using trained convolutional neural networks can meet the particular demands of fractional pixel interpolation.Use this hair
Bright obtained convolutional neural networks, which carry out fractional pixel interpolation, can obtain significant coding efficiency lifting, be more suitable for video volume
The fractional pixel interpolation part of code inter prediction.
The specific embodiment of the present invention is described above.It is to be appreciated that the invention is not limited in above-mentioned
Particular implementation, those skilled in the art can make various deformations or amendments within the scope of the claims, this not shadow
Ring the substantive content of the present invention.
Claims (6)
- A kind of 1. construction method of convolutional neural networks for Video coding fractional pixel interpolation, it is characterised in that:The side Method includes:Collect different content, the image of different resolution, formed comprising different type, the data of different coding complexity it is original Training dataset;Pretreatment operation is carried out to the original training data collection that is collected into, obtains meeting Video coding inter prediction fraction pixel and inserts It is worth the training data of characteristic, input data of the data as training convolutional neural networks;Depth convolutional neural networks are built, Video coding fractional pixel interpolation characteristic is considered, obtains being suitable for Video coding interframe Predict the convolutional neural networks structure of fractional pixel interpolation;The convolutional neural networks put up of data input obtained using pretreatment, at the same using the original training data collection as Corresponding true value, the convolutional neural networks that training is built, obtain the volume suitable for Video coding inter prediction fractional pixel interpolation Product neural network model.
- 2. the construction method of the convolutional neural networks according to claim 1 for Video coding fractional pixel interpolation, its It is characterized in that:The pretreatment operation, process are as follows:A) image that the fractional pixel position of interpolation generation concentrates original training data as needed carries out the drop of corresponding multiplying power Sampling operation, obtains the low resolution training data being used in step b);B) low resolution training data is encoded according to the configuration in video encoding standard to still image coding, is used Low resolution coding and rebuilding image in step c);C) up-sampling for carrying out corresponding to multiplying power to low resolution coding and rebuilding image in step a) operates, and returns to original image ruler It is very little, obtain the input data of training convolutional neural networks.
- 3. the construction method of the convolutional neural networks according to claim 2 for Video coding fractional pixel interpolation, its It is characterized in that:It is described c) in, the up-sampling of low resolution coding and rebuilding image is operated, ensures high resolution graphics after up-sampling As the pixel value of integer pixel positions is consistent with the low resolution coding and rebuilding figure before up-sampling.
- 4. it is used for the structure of the convolutional neural networks of Video coding fractional pixel interpolation according to claim 1-3 any one of them Method, it is characterised in that:It is described to build depth convolutional neural networks, wherein the depth convolutional neural networks built include 20 power Double-layer and 1 weight masking layer;For weight masking layer, WIFor the weight of integer pixel positions, WHFor fractional pixel position Weight, all fractional pixel positions share a weight.
- 5. the construction method of the convolutional neural networks according to claim 4 for Video coding fractional pixel interpolation, its It is characterized in that:The Video coding inter prediction fractional pixel interpolation, wherein integer pixel positions pixel value is constant, only generates point Number location of pixels.
- 6. the application for the convolutional neural networks model that a kind of any one of claim 1-5 the method is built, its feature exist In:The convolutional neural networks model is operated for fractional pixel interpolation, realizes the Video coding based on convolutional neural networks Inter prediction fractional pixel interpolation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711207766.7A CN108012157B (en) | 2017-11-27 | 2017-11-27 | Method for constructing convolutional neural network for video coding fractional pixel interpolation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711207766.7A CN108012157B (en) | 2017-11-27 | 2017-11-27 | Method for constructing convolutional neural network for video coding fractional pixel interpolation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108012157A true CN108012157A (en) | 2018-05-08 |
CN108012157B CN108012157B (en) | 2020-02-04 |
Family
ID=62054016
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711207766.7A Active CN108012157B (en) | 2017-11-27 | 2017-11-27 | Method for constructing convolutional neural network for video coding fractional pixel interpolation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108012157B (en) |
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109361919A (en) * | 2018-10-09 | 2019-02-19 | 四川大学 | A kind of image coding efficiency method for improving combined super-resolution and remove pinch effect |
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 |
CN109525859A (en) * | 2018-10-10 | 2019-03-26 | 腾讯科技(深圳)有限公司 | Model training, image transmission, image processing method and relevant apparatus equipment |
CN109785279A (en) * | 2018-12-28 | 2019-05-21 | 江苏师范大学 | A kind of image co-registration method for reconstructing based on deep learning |
CN110072119A (en) * | 2019-04-11 | 2019-07-30 | 西安交通大学 | A kind of perception of content video adaptive transmission method based on deep learning network |
CN110099280A (en) * | 2019-05-24 | 2019-08-06 | 浙江大学 | A kind of video service quality Enhancement Method under wireless self-organization network Bandwidth-Constrained |
CN110177282A (en) * | 2019-05-10 | 2019-08-27 | 杭州电子科技大学 | A kind of inter-frame prediction method based on SRCNN |
CN110493596A (en) * | 2019-09-02 | 2019-11-22 | 西北工业大学 | A kind of video coding framework neural network based |
CN110502954A (en) * | 2018-05-17 | 2019-11-26 | 杭州海康威视数字技术股份有限公司 | The method and apparatus of video analysis |
CN110519606A (en) * | 2019-08-22 | 2019-11-29 | 天津大学 | Intelligent coding method in a kind of deep video frame |
CN110572710A (en) * | 2019-09-25 | 2019-12-13 | 北京达佳互联信息技术有限公司 | video generation method, device, equipment and storage medium |
CN110794254A (en) * | 2018-08-01 | 2020-02-14 | 北京映翰通网络技术股份有限公司 | Power distribution network fault prediction method and system based on reinforcement learning |
CN110794255A (en) * | 2018-08-01 | 2020-02-14 | 北京映翰通网络技术股份有限公司 | Power distribution network fault prediction method and system |
CN110933432A (en) * | 2018-09-19 | 2020-03-27 | 珠海金山办公软件有限公司 | Image compression method, image decompression method, image compression device, image decompression device, electronic equipment and storage medium |
CN111010568A (en) * | 2018-10-06 | 2020-04-14 | 华为技术有限公司 | Interpolation filter training method and device, video image coding and decoding method and coder and decoder |
CN111711817A (en) * | 2019-03-18 | 2020-09-25 | 四川大学 | HEVC intra-frame coding compression performance optimization research combined with convolutional neural network |
CN111800630A (en) * | 2019-04-09 | 2020-10-20 | Tcl集团股份有限公司 | Method and system for reconstructing video super-resolution and electronic equipment |
CN112601095A (en) * | 2020-11-19 | 2021-04-02 | 北京影谱科技股份有限公司 | Method and system for creating fractional interpolation model of video brightness and chrominance |
CN112889283A (en) * | 2018-10-19 | 2021-06-01 | 三星电子株式会社 | Encoding method and apparatus thereof, and decoding method and apparatus thereof |
CN113365079A (en) * | 2021-06-01 | 2021-09-07 | 闽南师范大学 | Video coding pixel motion compensation method based on super-resolution network |
CN113822801A (en) * | 2021-06-28 | 2021-12-21 | 浙江工商大学 | Compressed video super-resolution reconstruction method based on multi-branch convolutional neural network |
CN114677652A (en) * | 2022-05-30 | 2022-06-28 | 武汉博观智能科技有限公司 | Illegal behavior monitoring method and device |
US11445198B2 (en) * | 2020-09-29 | 2022-09-13 | Tencent America LLC | Multi-quality video super resolution with micro-structured masks |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104112263A (en) * | 2014-06-28 | 2014-10-22 | 南京理工大学 | Method for fusing full-color image and multispectral image based on deep neural network |
WO2016132145A1 (en) * | 2015-02-19 | 2016-08-25 | Magic Pony Technology Limited | Online training of hierarchical algorithms |
CN106204449A (en) * | 2016-07-06 | 2016-12-07 | 安徽工业大学 | A kind of single image super resolution ratio reconstruction method based on symmetrical degree of depth network |
-
2017
- 2017-11-27 CN CN201711207766.7A patent/CN108012157B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104112263A (en) * | 2014-06-28 | 2014-10-22 | 南京理工大学 | Method for fusing full-color image and multispectral image based on deep neural network |
WO2016132145A1 (en) * | 2015-02-19 | 2016-08-25 | Magic Pony Technology Limited | Online training of hierarchical algorithms |
CN106204449A (en) * | 2016-07-06 | 2016-12-07 | 安徽工业大学 | A kind of single image super resolution ratio reconstruction method based on symmetrical degree of depth network |
Cited By (39)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110502954A (en) * | 2018-05-17 | 2019-11-26 | 杭州海康威视数字技术股份有限公司 | The method and apparatus of video analysis |
CN110794255A (en) * | 2018-08-01 | 2020-02-14 | 北京映翰通网络技术股份有限公司 | Power distribution network fault prediction method and system |
CN110794255B (en) * | 2018-08-01 | 2022-01-18 | 北京映翰通网络技术股份有限公司 | Power distribution network fault prediction method and system |
CN110794254A (en) * | 2018-08-01 | 2020-02-14 | 北京映翰通网络技术股份有限公司 | Power distribution network fault prediction method and system based on reinforcement learning |
CN110933432A (en) * | 2018-09-19 | 2020-03-27 | 珠海金山办公软件有限公司 | Image compression method, image decompression method, image compression device, image decompression device, electronic equipment and storage medium |
CN111010568A (en) * | 2018-10-06 | 2020-04-14 | 华为技术有限公司 | Interpolation filter training method and device, video image coding and decoding method and coder and decoder |
CN111010568B (en) * | 2018-10-06 | 2023-09-29 | 华为技术有限公司 | Training method and device of interpolation filter, video image coding and decoding method and coder-decoder |
CN109361919A (en) * | 2018-10-09 | 2019-02-19 | 四川大学 | A kind of image coding efficiency method for improving combined super-resolution and remove pinch effect |
CN109525859B (en) * | 2018-10-10 | 2021-01-15 | 腾讯科技(深圳)有限公司 | Model training method, image sending method, image processing method and related device equipment |
CN109525859A (en) * | 2018-10-10 | 2019-03-26 | 腾讯科技(深圳)有限公司 | Model training, image transmission, image processing method and relevant apparatus equipment |
CN112889283A (en) * | 2018-10-19 | 2021-06-01 | 三星电子株式会社 | Encoding method and apparatus thereof, and decoding method and apparatus thereof |
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 |
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 |
CN109785279A (en) * | 2018-12-28 | 2019-05-21 | 江苏师范大学 | A kind of image co-registration method for reconstructing based on deep learning |
CN109785279B (en) * | 2018-12-28 | 2023-02-10 | 江苏师范大学 | Image fusion reconstruction method based on deep learning |
CN111711817B (en) * | 2019-03-18 | 2023-02-10 | 四川大学 | HEVC intra-frame coding compression performance optimization method combined with convolutional neural network |
CN111711817A (en) * | 2019-03-18 | 2020-09-25 | 四川大学 | HEVC intra-frame coding compression performance optimization research combined with convolutional neural network |
CN111800630A (en) * | 2019-04-09 | 2020-10-20 | Tcl集团股份有限公司 | Method and system for reconstructing video super-resolution and electronic equipment |
CN110072119B (en) * | 2019-04-11 | 2020-04-10 | 西安交通大学 | Content-aware video self-adaptive transmission method based on deep learning network |
CN110072119A (en) * | 2019-04-11 | 2019-07-30 | 西安交通大学 | A kind of perception of content video adaptive transmission method based on deep learning network |
CN110177282A (en) * | 2019-05-10 | 2019-08-27 | 杭州电子科技大学 | A kind of inter-frame prediction method based on SRCNN |
CN110177282B (en) * | 2019-05-10 | 2021-06-04 | 杭州电子科技大学 | Interframe prediction method based on SRCNN |
CN110099280B (en) * | 2019-05-24 | 2020-05-08 | 浙江大学 | Video service quality enhancement method under limitation of wireless self-organizing network bandwidth |
CN110099280A (en) * | 2019-05-24 | 2019-08-06 | 浙江大学 | A kind of video service quality Enhancement Method under wireless self-organization network Bandwidth-Constrained |
CN110519606A (en) * | 2019-08-22 | 2019-11-29 | 天津大学 | Intelligent coding method in a kind of deep video frame |
CN110519606B (en) * | 2019-08-22 | 2021-12-07 | 天津大学 | Depth video intra-frame intelligent coding method |
CN110493596B (en) * | 2019-09-02 | 2021-09-17 | 西北工业大学 | Video coding system and method based on neural network |
CN110493596A (en) * | 2019-09-02 | 2019-11-22 | 西北工业大学 | A kind of video coding framework neural network based |
CN110572710A (en) * | 2019-09-25 | 2019-12-13 | 北京达佳互联信息技术有限公司 | video generation method, device, equipment and storage medium |
CN110572710B (en) * | 2019-09-25 | 2021-09-28 | 北京达佳互联信息技术有限公司 | Video generation method, device, equipment and storage medium |
US11445198B2 (en) * | 2020-09-29 | 2022-09-13 | Tencent America LLC | Multi-quality video super resolution with micro-structured masks |
CN112601095B (en) * | 2020-11-19 | 2023-01-10 | 北京影谱科技股份有限公司 | Method and system for creating fractional interpolation model of video brightness and chrominance |
CN112601095A (en) * | 2020-11-19 | 2021-04-02 | 北京影谱科技股份有限公司 | Method and system for creating fractional interpolation model of video brightness and chrominance |
CN113365079A (en) * | 2021-06-01 | 2021-09-07 | 闽南师范大学 | Video coding pixel motion compensation method based on super-resolution network |
CN113365079B (en) * | 2021-06-01 | 2023-05-30 | 闽南师范大学 | Super-resolution network-based video coding sub-pixel motion compensation method |
CN113822801B (en) * | 2021-06-28 | 2023-08-18 | 浙江工商大学 | Compressed video super-resolution reconstruction method based on multi-branch convolutional neural network |
CN113822801A (en) * | 2021-06-28 | 2021-12-21 | 浙江工商大学 | Compressed video super-resolution reconstruction method based on multi-branch convolutional neural network |
CN114677652A (en) * | 2022-05-30 | 2022-06-28 | 武汉博观智能科技有限公司 | Illegal behavior monitoring method and device |
Also Published As
Publication number | Publication date |
---|---|
CN108012157B (en) | 2020-02-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108012157A (en) | Construction method for the convolutional neural networks of Video coding fractional pixel interpolation | |
Lai et al. | Deep laplacian pyramid networks for fast and accurate super-resolution | |
Shi et al. | Single image super-resolution with dilated convolution based multi-scale information learning inception module | |
CN104574336B (en) | Super-resolution image reconstruction system based on adaptive sub- mould dictionary selection | |
Wang et al. | Contextual transformation network for lightweight remote-sensing image super-resolution | |
CN109064396A (en) | A kind of single image super resolution ratio reconstruction method based on depth ingredient learning network | |
CN110087092A (en) | Low bit-rate video decoding method based on image reconstruction convolutional neural networks | |
CN109410146A (en) | A kind of image deblurring algorithm based on Bi-Skip-Net | |
DE202012013410U1 (en) | Image compression with SUB resolution images | |
US11328184B2 (en) | Image classification and conversion method and device, image processor and training method therefor, and medium | |
CN104657962B (en) | The Image Super-resolution Reconstruction method returned based on cascading linear | |
CN104199627B (en) | Gradable video encoding system based on multiple dimensioned online dictionary learning | |
CN108900848A (en) | A kind of video quality Enhancement Method based on adaptive separable convolution | |
Wang et al. | Lightweight single image super-resolution convolution neural network in portable device. | |
CN105049851A (en) | Channel no-reference image quality evaluation method based on color perception | |
CN110533591B (en) | Super-resolution image reconstruction method based on codec structure | |
CN113079378B (en) | Image processing method and device and electronic equipment | |
DE102014115013A1 (en) | Video coding method and apparatus, and video decoding method and apparatus performing motion compensation | |
CN104537610A (en) | Super-resolution image reconstruction method based on Sparse representation and UV channel processing | |
CN117560511A (en) | Spacer image compression method and system based on graph segmentation technology and electric power inspection | |
CN112906874A (en) | Convolutional neural network characteristic graph data compression method and device | |
CN106157251A (en) | A kind of face super-resolution method based on Cauchy's regularization | |
Xin et al. | FISTA-CSNet: a deep compressed sensing network by unrolling iterative optimization algorithm | |
Jeevan et al. | WaveMixSR: Resource-efficient neural network for image super-resolution | |
CN109672891A (en) | The lossless second-compressed method of jpeg image |
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 |