CN109361919A - A kind of image coding efficiency method for improving combined super-resolution and remove pinch effect - Google Patents
A kind of image coding efficiency method for improving combined super-resolution and remove pinch effect Download PDFInfo
- Publication number
- CN109361919A CN109361919A CN201811170786.6A CN201811170786A CN109361919A CN 109361919 A CN109361919 A CN 109361919A CN 201811170786 A CN201811170786 A CN 201811170786A CN 109361919 A CN109361919 A CN 109361919A
- Authority
- CN
- China
- Prior art keywords
- resolution
- image
- coding
- super
- coding mode
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 95
- 230000000694 effects Effects 0.000 title claims abstract description 51
- 230000006835 compression Effects 0.000 claims abstract description 24
- 238000007906 compression Methods 0.000 claims abstract description 24
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 5
- 238000003908 quality control method Methods 0.000 claims abstract description 4
- 230000003044 adaptive effect Effects 0.000 claims abstract description 3
- 230000006837 decompression Effects 0.000 claims description 36
- 238000012549 training Methods 0.000 claims description 4
- 230000000295 complement effect Effects 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 17
- 238000012545 processing Methods 0.000 description 15
- 238000002474 experimental method Methods 0.000 description 13
- 238000005070 sampling Methods 0.000 description 13
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 6
- 230000006872 improvement Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000004927 fusion Effects 0.000 description 4
- 230000007704 transition Effects 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000000903 blocking effect Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000011056 performance test Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000003321 amplification Effects 0.000 description 2
- 238000012733 comparative method Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000003199 nucleic acid amplification method Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 101100001677 Emericella variicolor andL gene Proteins 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
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
- H04N19/103—Selection of coding mode or of prediction mode
-
- 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
- H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a kind of image coding efficiency method for improving combined super-resolution and remove pinch effect.It mainly comprises the steps that for jpeg compressed image, construct and trains the depth convolutional neural networks for removing pinch effect and super-resolution；Based on given image and coding quality control parameter to be encoded, coding mode is selected using adaptive coding mode adjudging method；Based on selected coding mode, given image is encoded.Based on super-resolution and pinch effect technology is gone, the method for the present invention provides low resolution coding mode and full resolution coding mode simultaneously to be adapted to the image of different characteristics and different compression ratios.The method of the invention can effectively improve the coding efficiency of JPEG, and it can be extended in other images and video encoding standard.The method of the present invention can be applied to the fields such as image coding and Video coding.
Description
Technical Field
The invention relates to an image compression coding technology, in particular to an image coding performance improving method combining super-resolution and decompression effects, and belongs to the field of image processing.
Background
With the rapid development and popularization of imaging devices and display devices, images and videos become the main information carriers. The problem with this is that the amount of image and video data increases dramatically, which presents a significant challenge to existing storage and transmission systems. On the other hand, the resolution of images and videos is increasing, so that the pressure on storage and transmission systems is further increasing. Lossy coding techniques can significantly reduce the amount of data at the expense of some degree of signal distortion by reducing redundancy of the data. However, due to the loss of information during the encoding process, there are often varying degrees of compression effects in the decoded signal. When storage and transmission resources are limited, the compression effect is particularly serious, and the subjective quality and the utilization value of an image signal are seriously influenced. Therefore, there is a need to further improve the performance of existing compression coding algorithms to accommodate the rapidly increasing amount of image and video signal data.
Disclosure of Invention
The invention aims to provide a method for improving the self-adaptive coding performance aiming at the existing image coding standard.
The invention provides an image coding performance improving method combining super-resolution and decompression effects, which mainly comprises the following operation steps:
the method comprises the following steps of (I) constructing and training a deep convolutional neural network for a decompression effect and super-resolution aiming at a JPEG compressed image;
based on the given image to be coded and the coding quality control parameter, selecting a coding mode by using a self-adaptive coding mode judgment method;
and (III) coding the given image based on the selected coding mode.
Drawings
FIG. 1 is a block diagram of an image coding performance improving method combining super-resolution and decompression effects according to the present invention
FIG. 2 is a block diagram of the method for removing the effect of compression based on the enhanced residual error module in the present invention
FIG. 3 is a block diagram of a super-resolution method based on an enhanced residual module in the present invention
FIG. 4 is a block diagram of an enhanced residual module in the present invention
FIG. 5 is a graph of the fit between the sampling distortion and the mode transition threshold in the present invention
Fig. 6 shows the processing result of the decompression effect removing method and the comparison method in the present invention on the test image "barbarba" (in this experiment, the JPEG quality factor is set to 10): wherein, (a) is the original test image, (b) (c) (d) (e) are the processing results of JPEG, comparison method 1, comparison method 2 and the decompression effect method in the invention
Fig. 7 shows the processing result of the test image "img 092" by the super-resolution method and the comparison method in the present invention (in this experiment, the JPEG quality factor is set to 10, and the super-resolution factor is set to 2): wherein, (a) is the original test image, (b) (c) (d) (e) are bicubic interpolation, comparison method 3, comparison method 4 and the processing result of the super-resolution method in the invention
Fig. 8 is a comparison of the coding performance improvement method and the rate-distortion performance of JPEG in the present invention: wherein the test images of (a) and (b) are "Kodim 05" and "Kodim 07", respectively "
Fig. 9 is a decoding image of the JPEG pair test image "Kodim 05" and the encoding performance improvement method in the present invention: wherein, (a) is original test image, (b) is decoding image of JPEG with code rate of 0.197bpp, (c) is decoding image of coding performance improving method with code rate of 0.196bpp, (d) is decoding image of JPEG with code rate of 0.512bpp, and (e) is decoding image of coding performance improving method with code rate of 0.512bpp
Fig. 10 is a decoding image of the JPEG pair test image "Kodim 07" and the encoding performance improvement method in the present invention: wherein, (a) is original test image, (b) is decoding image of JPEG with code rate of 0.200bpp, (c) is decoding image of coding performance improving method with code rate of 0.199bpp, (d) is decoding image of JPEG with code rate of 0.500bpp, and (e) is decoding image of coding performance improving method with code rate of 0.500bpp
Detailed Description
The invention will be further described with reference to the accompanying drawings in which:
in fig. 1, a method for improving image coding performance by combining super-resolution and decompression effects may specifically include the following three steps:
the method comprises the following steps of (I) constructing and training a deep convolutional neural network for a decompression effect and super-resolution aiming at a JPEG compressed image;
based on the given image to be coded and the coding quality control parameter, selecting a coding mode by using a self-adaptive coding mode judgment method;
and (III) coding the given image based on the selected coding mode.
Specifically, in the step (one), a decompression effect network shown in fig. 2 and a super-resolution network shown in fig. 3 are constructed, and the structure of the enhanced residual module in fig. 2 and 3 is shown in fig. 4.
The basic building blocks of the enhanced residual block shown on the right side of fig. 4 are the residual blocks on the left side. The residual module is composed of convolutional layer, nonlinear activation function, convolutional layer, and is expressed as
WhereinAndrespectively representing the input and output of the residual module;andrespectively representing the first and second convolution layers in the residual module (in the invention, the number of convolution kernels of the convolution layers in the residual module is set to be 64, and the size is set to be 3 multiplied by 3); σ (-) denotes the nonlinear activation layer, which is defined as max (0,). For ease of representation, the mapping of the residual block implementation is denoted as FRB(. a) is namely
The enhanced residual module in the present invention is mainly composed of five residual modules, which are denoted as
WhereinAndrespectively representing the input and the output of the enhanced residual error module;represents the ith residual block in the enhanced residual block,and is provided with[·]Representing a feature join operation (feature join layer);representing the feature fusion operation in the enhanced residual module (feature fusion layer, number of convolution kernels set to 64, size set to 1 × 1);represents the linear transform operation (linear transform layer) in the enhanced residual module. For ease of representation, the mapping of the enhanced residual block implementation is denoted as FERB(. a) is namely
The basic elements of the decompression effect network shown in fig. 2 are the enhanced residual module in fig. 4. Specifically, the input image I with the decompression effect is first decompressedARDecomposed into four sub-pictures (sampling layers) of low resolution, i.e. having
Wherein M and N represent the original input image I, respectivelyARHeight and width of (a). Further, four sub-graphs obtained by decomposition are connected and combined into a four-channel tensorNamely have
For theFirst, initial features are extracted using a convolutional layer, i.e.
Wherein,represents the feature extraction operation in the decompression effect network (feature extraction layer, number of convolution kernels set to 64, size set to 3 × 3).
The initial features are then transformed step by step using four enhanced residual modules, i.e.
Wherein, FERB,iRepresents the ith enhanced residual error module in the decompression effect network and has
Furthermore, the features of different layers generated by the enhanced residual modules are fused, and the residual components are reconstructed. Reconstructed four-channel tensorIs shown as
Wherein,representing the feature fusion operation in the decompression effect network (feature fusion layer, number of convolution kernels is set to 64, size is set to 1 × 1);representing residual reconstruction operations in the decompression effect network (residual reconstruction layers, number of convolution kernels set to 4, size set to 3 × 3);represents the linear transformation operation (linear transformation layer) in the decompression effect network.
Finally, the result is processed by removing the compression effectByAre combined (recombination layer), i.e.
The super-resolution network shown in fig. 3 is basically the same as the decompression effect network structure shown in fig. 2, and the main difference is that an interpolation and enlargement operation (interpolation and enlargement layer) is introduced into the super-resolution network. The input of the super-resolution network is a low-resolution image, and the output is a high-resolution image. Therefore, the main purpose of the interpolation magnification operation is to reconstruct the original high resolution image. Except for the interpolation amplification layer, the structures and parameter settings of the super-resolution network and the decompression effect network are completely consistent. In the super-resolution network shown in fig. 3, interpolation amplification is realized by bicubic interpolation. For both the decompression effect network shown in fig. 2 and the super-resolution network shown in fig. 3, the present invention adopts a mean square error cost function for training.
Specifically, in the step (two), for a given image to be encoded, we select a suitable mode from two alternative encoding modes by an adaptive encoding mode decision method, where the alternative modes include a full resolution encoding mode and a low resolution encoding mode. The details of the full resolution coding mode and the low resolution coding mode will be given in step three. Overall, the full resolution coding mode is more suitable for the middle and high rate segments, and the low resolution coding mode is more suitable for the low rate segments. In the present invention, the optimal encoding mode is related to image characteristics and encoding quality parameters. Considering that the main difference between the full resolution coding mode and the low resolution coding mode is the sampling link, we use sampling distortion to characterize the image. In particular, for a given image I, the sampling distortion is defined as
Wherein IduThe sampling factor is 2 for the result of I being successively down-sampled and up-sampled. I | · | purple wind1Is the norm of L1, and T is the number of pixels in I.
In view of adjusting the compression rate and the code rate by the Quality Factor (QF) in JPEG, the present invention establishes a relationship between the sampling distortion and the threshold QF of the encoding mode conversion using the sample image. The threshold QF for the encoding mode transition is defined as the critical QF for the full resolution encoding mode and the low resolution encoding mode transition. Fig. 5 shows a conversion relationship between sample points and fitting, wherein the horizontal axis represents sampling distortion and the vertical axis represents the encoding mode conversion threshold QF. Based on the distribution of the samples, the sample points are fitted with power functions, i.e.
q＝AdB(10)
Wherein q is a critical threshold QF for switching two coding modes, and A and B are fitting parameters. In the present invention, a is 24.55 and B is-0.5258.
For an image I and a given JPEG coding quality factor QIFirst, the sampling distortion d is calculated according to the formula (9)IThen, the threshold q of the mode transition is estimated by a fitting function (10)I. When Q isIGreater than or equal to qIThen, a full resolution coding mode is adopted; when Q isILess than qIThe low resolution coding mode is used.
Specifically, in the step (three), when the selected coding mode is the full-resolution coding mode, the existing standard codec (such as JPEG) is firstly adopted to encode and decode the image to be encoded; then, the decoded image is processed by using the decompression effect network trained in the step (one) of the invention to suppress the compression noise. The full resolution coding mode is represented as
If＝Ar(Dc(Ec(I))) (11)
Wherein, I and IfRespectively representing an input image and an output image in a full resolution encoding mode; ec(·)、Dc(. and A)r(. cndot.) represents the encoding, decoding, and decompression effects processes, respectively.
When the selected coding mode is a low-resolution coding mode, sampling an image I to be coded, and coding and decoding the sampled low-resolution image by adopting an existing standard coder-decoder; and finally, restoring the decoded low-resolution image to the original resolution by using the super-resolution network trained in the step (I) of the invention. The low resolution coding mode is represented as
Il＝Sr(Dc(Ec(Ds(I)))) (12)
Wherein, IlRepresenting an output image in a low resolution encoding mode; ds(. and S)r(. cndot.) denotes the down-sampling and super-resolution processes, respectively.
In order to test the performance of the decompression effect method, the super-resolution method and the coding performance improvement method, a decompression effect experiment, a super-resolution experiment and a coding performance test experiment are performed on a common test image.
For the decompression effect experiments, the test image was "Barbara" and the quality factor for JPEG compression was set to 10. Two comparative methods are:
the method comprises the following steps: the methods proposed by Zhang et al, references "J.Zhang, R.Xiong, C.Zhao, Y.Zhang, S.Ma, and W.Gao, CONCOLOR: Constrained non-Constrained low-rank model for image deblocking, IEEETranss.image Process, vol.25, No.3, pp.1246-1259,2016.
The method 2 comprises the following steps: the methods proposed by Zhang et al, references "K.Zhang, W.Zuo, Y.Chen, D.Meng, andL.Zhang, Beyond a gaussian denoiser: reactive learning of deep CNN for imaging modeling. image processing, vol.26, No.7, pp.3142-3155,2017.
For the super-resolution experiment, the test image is "img 092", the reconstruction factor is set to 2, and the quality factor for JPEG compression is set to 10. Two comparative methods are:
the method 3 comprises the following steps: methods proposed by Li et al, references "t.li, x.he, l.qing, q.teng, and h.chen, interactive frame of captured deblocking and preprocessing for compressed images, ieee trans.multimedia, vol.20, No.6, pp.1305-1320,2018.
The method 4 comprises the following steps: the method proposed by Kim et al, references "J.Kim, J.Kwon Lee, and K.mu Lee, accurate prediction using version restriction networks, proceedings of the IEEEConference on Computer Vision and Pattern Recognition (CVPR),2016, pp.1646-1654.
For the experiments of the coding performance test, the test images were "Kodim 05" and "Kodim 07". The coder and the decoder adopt JPEG, and the comparison method is JPEG.
Decompression effect experiment: the test image "Barbara" was compressed using JPEG with a quality factor of 10. The compression result is the input image of the decompression effect method. Fig. 6(a), 6(b), 6(c), 6(d), and 6(e) show the "barbarbara" original image, the JPEG compression result, and the processing results of the methods, respectively. The table shows objective evaluation parameters of the results of the methods in the experiment.
Watch 1
As can be seen from the experimental results in fig. 6, when the quality factor is 10, a very serious compression effect exists in the JPEG compressed image, wherein the blocking effect is particularly significant, and the subjective visual effect is seriously affected; the method 1, the method 2 and the method for removing the compression effect can better inhibit the compression effect and improve the image quality. However, as can be seen from the enlarged view, the method for removing the compression effect in the present invention can better maintain and restore the local structure while suppressing the compression effect. From the PSNR and SSIM parameters given in table one, the decompression effect method of the present invention is significantly better than JPEG, method 1 and method 2. In summary, the decompression effect method of the invention can effectively process the compression effect in the JPEG compressed image.
Super-resolution experiments: and performing 2-time down-sampling on the test image img092 by utilizing bicubic interpolation, and then compressing a sampling result by adopting JPEG (joint photographic experts group) with a compression quality factor of 10. The compression result is the input image of the super-resolution method. Fig. 7(a), 7(b), 7(c), 7(d), and 7(e) show the original image and the processing results of the methods, respectively. And the second table is objective evaluation parameters of processing results of each method in the experiment.
Watch two
As can be seen from the experimental results in fig. 7, the bicubic interpolation method has serious aliasing, blocking artifacts, and other distortions. The method 3 and the method 4 can inhibit partial blocking effect while improving the resolution, but the processing result still has obvious artificial effect; meanwhile, the processing results of the methods 3 and 4 are also blurred to some extent. In comparison, the super-resolution method of the invention better inhibits the compression effect, the processing result is clearer, and the image edge is sharper. As can be seen from the second table, the PSNR and SSIM of the processing results of the super-resolution method in the invention are obviously superior to those of the comparison method. In summary, the super-resolution method of the invention can effectively realize the super-resolution of JPEG compressed images.
Coding performance test experiment: under different compression rates, the test images 'Kodim 05' and 'Kodim 07' are encoded by JPEG and the encoding performance improving method in the invention. Fig. 8 shows rate-distortion curves of JPEG and the coding performance improvement method of the present invention. The "Kodim 05" original image and the processing results of the methods are shown in fig. 9. The "Kodim 07" original image and the processing results of the methods are shown in fig. 10.
As can be seen from the rate-distortion curve shown in fig. 8, the rate-distortion performance of the coding performance improving method of the present invention is comprehensively superior to JPEG. Under the same code rate, PSNR is obviously improved; moreover, the effective code rate range of the coding performance improving method basically covers the commonly used code rate segment. The results in fig. 9 and fig. 10 show that, at the same code rate, the decoding result of the coding performance improving method in the present invention has a better visual effect than the JPEG result. In summary, the coding performance improving method of the invention can effectively improve the coding performance of JPEG.
Claims (6)
1. A method for improving image coding performance by combining super-resolution and decompression effects is characterized by comprising the following steps:
the method comprises the following steps: constructing and training a deep convolutional neural network for decompression effect and super-resolution aiming at a JPEG compressed image;
step two: selecting a coding mode by using a self-adaptive coding mode judgment method based on a given image to be coded and a coding quality control parameter;
step three: a given image is encoded based on the selected encoding mode.
2. The method for improving image coding performance combining super-resolution and decompression effect as claimed in claim 1, wherein the method for improving decompression effect based on enhanced residual error module in step one.
3. The method for improving image coding performance combining super resolution and decompression effect according to claim 1, wherein the super resolution method based on enhanced residual error module in step one.
4. The method for improving image coding performance combining super-resolution and decompression effect according to claim 1, wherein the adaptive coding mode decision method based on statistical modeling in step two is adopted.
5. The method of claim 1, wherein two complementary coding modes in the third step are full resolution coding mode and low resolution coding mode.
6. The method for improving image coding performance combining super-resolution and decompression effects according to claims 1-5, which can be extended to other image or video compression standards.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811170786.6A CN109361919A (en) | 2018-10-09 | 2018-10-09 | A kind of image coding efficiency method for improving combined super-resolution and remove pinch effect |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811170786.6A CN109361919A (en) | 2018-10-09 | 2018-10-09 | A kind of image coding efficiency method for improving combined super-resolution and remove pinch effect |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109361919A true CN109361919A (en) | 2019-02-19 |
Family
ID=65348567
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811170786.6A Pending CN109361919A (en) | 2018-10-09 | 2018-10-09 | A kind of image coding efficiency method for improving combined super-resolution and remove pinch effect |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109361919A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109996071A (en) * | 2019-03-27 | 2019-07-09 | 上海交通大学 | Variable bit rate image coding, decoding system and method based on deep learning |
| CN110222754A (en) * | 2019-05-28 | 2019-09-10 | 天津大学 | Full resolution depth convolutional neural networks image classification method based on tensor analysis |
| CN110401834A (en) * | 2019-08-06 | 2019-11-01 | 杭州微帧信息科技有限公司 | A kind of adaptive video coding method based on deep learning |
| CN111726638A (en) * | 2019-03-18 | 2020-09-29 | 四川大学 | HEVC (high efficiency video coding) optimization method combining decompression effect and super-resolution |
| CN113225590A (en) * | 2021-05-06 | 2021-08-06 | 深圳思谋信息科技有限公司 | Video super-resolution enhancement method and device, computer equipment and storage medium |
| CN113837935A (en) * | 2020-06-24 | 2021-12-24 | 四川大学 | Compressed image super-resolution reconstruction method based on attention-enhancing network |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016132146A1 (en) * | 2015-02-19 | 2016-08-25 | Magic Pony Technology Limited | Visual processing using sub-pixel convolutions |
| CN107018422A (en) * | 2017-04-27 | 2017-08-04 | 四川大学 | Still image compression method based on depth convolutional neural networks |
| CN107563965A (en) * | 2017-09-04 | 2018-01-09 | 四川大学 | Jpeg compressed image super resolution ratio reconstruction method based on convolutional neural networks |
| CN108012157A (en) * | 2017-11-27 | 2018-05-08 | 上海交通大学 | Construction method for the convolutional neural networks of Video coding fractional pixel interpolation |
| CN108495128A (en) * | 2018-03-09 | 2018-09-04 | 罗雷 | Video transmission system based on super-resolution rebuilding technology and transmission method |
-
2018
- 2018-10-09 CN CN201811170786.6A patent/CN109361919A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016132146A1 (en) * | 2015-02-19 | 2016-08-25 | Magic Pony Technology Limited | Visual processing using sub-pixel convolutions |
| CN107018422A (en) * | 2017-04-27 | 2017-08-04 | 四川大学 | Still image compression method based on depth convolutional neural networks |
| CN107563965A (en) * | 2017-09-04 | 2018-01-09 | 四川大学 | Jpeg compressed image super resolution ratio reconstruction method based on convolutional neural networks |
| CN108012157A (en) * | 2017-11-27 | 2018-05-08 | 上海交通大学 | Construction method for the convolutional neural networks of Video coding fractional pixel interpolation |
| CN108495128A (en) * | 2018-03-09 | 2018-09-04 | 罗雷 | Video transmission system based on super-resolution rebuilding technology and transmission method |
Non-Patent Citations (3)
| Title |
|---|
| XIAOHAI HE; HONGGANG CHEN; JINGXU CHEN; LINBO QING: "Soft decoding of JPEG 2000 compressed images using bit-rate-driven deep convolutional neural networks", 《 2017 IEEE INTERNATIONAL CONFERENCE ON INFORMATION AND AUTOMATION (ICIA)》 * |
| 何小海；吴笛；滕奇志；卿粼波；黄建秋: "集成超分辨率重建的图像压缩编码新型框架及其实现", 《数据采集与处理》 * |
| 戴茂华；王正勇；卿粼波；何小海: "超分辨率重建在低码率视频压缩编码中的应用", 《电视技术》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111726638A (en) * | 2019-03-18 | 2020-09-29 | 四川大学 | HEVC (high efficiency video coding) optimization method combining decompression effect and super-resolution |
| CN109996071A (en) * | 2019-03-27 | 2019-07-09 | 上海交通大学 | Variable bit rate image coding, decoding system and method based on deep learning |
| CN109996071B (en) * | 2019-03-27 | 2020-03-27 | 上海交通大学 | Variable code rate image coding and decoding system and method based on deep learning |
| CN110222754A (en) * | 2019-05-28 | 2019-09-10 | 天津大学 | Full resolution depth convolutional neural networks image classification method based on tensor analysis |
| CN110222754B (en) * | 2019-05-28 | 2023-06-06 | 天津大学 | Full-resolution depth convolution neural network image classification method based on tensor analysis |
| CN110401834A (en) * | 2019-08-06 | 2019-11-01 | 杭州微帧信息科技有限公司 | A kind of adaptive video coding method based on deep learning |
| CN113837935A (en) * | 2020-06-24 | 2021-12-24 | 四川大学 | Compressed image super-resolution reconstruction method based on attention-enhancing network |
| CN113225590A (en) * | 2021-05-06 | 2021-08-06 | 深圳思谋信息科技有限公司 | Video super-resolution enhancement method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109361919A (en) | A kind of image coding efficiency method for improving combined super-resolution and remove pinch effect | |
| CN110087092B (en) | Low-bit-rate video coding and decoding method based on image reconstruction convolutional neural network | |
| CN109495741B (en) | Image compression method based on self-adaptive down-sampling and deep learning | |
| CN110751597B (en) | Video super-resolution method based on coding damage repair | |
| CN111711817B (en) | HEVC intra-frame coding compression performance optimization method combined with convolutional neural network | |
| CN103607591A (en) | Image compression method combining super-resolution reconstruction | |
| CN111885280B (en) | Hybrid convolutional neural network video coding loop filtering method | |
| WO2023000179A1 (en) | Video super-resolution network, and video super-resolution, encoding and decoding processing method and device | |
| CN107181949A (en) | A kind of compression of images framework of combination super-resolution and residual coding technology | |
| CN111726614A (en) | HEVC (high efficiency video coding) optimization method based on spatial domain downsampling and deep learning reconstruction | |
| CN111726638A (en) | HEVC (high efficiency video coding) optimization method combining decompression effect and super-resolution | |
| CN109922339A (en) | In conjunction with the image coding framework of multi-sampling rate down-sampling and super-resolution rebuilding technology | |
| CN113592746A (en) | Method for enhancing quality of compressed video by fusing space-time information from coarse to fine | |
| CN112218094A (en) | JPEG image decompression effect removing method based on DCT coefficient prediction | |
| CN111669588B (en) | Ultra-high definition video compression coding and decoding method with ultra-low time delay | |
| CN115150628B (en) | Method for coding coarse-to-fine depth video with super-priori guided mode prediction | |
| CN112188217A (en) | JPEG compressed image decompression effect removing method combining DCT domain and pixel domain learning | |
| Tan et al. | Image compression algorithms based on super-resolution reconstruction technology | |
| Abeda et al. | Pixel based techniques for gray image compression: A review | |
| Han et al. | HDR image compression with convolutional autoencoder | |
| Jia et al. | Deep convolutional network based image quality enhancement for low bit rate image compression | |
| CN113709483A (en) | Adaptive generation method and device for interpolation filter coefficient | |
| Nasiri et al. | A study on the impact of training data in CNN-based super-resolution for low bitrate end-to-end video coding | |
| CN115358954B (en) | Attention-guided feature compression method | |
| Kamble et al. | Quadtree partitioning and extended weighted finite automata-based fractal colour video coding |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190219 |