CN102930518A - Improved sparse representation based image super-resolution method - Google Patents
Improved sparse representation based image super-resolution method Download PDFInfo
- Publication number
- CN102930518A CN102930518A CN2012101956206A CN201210195620A CN102930518A CN 102930518 A CN102930518 A CN 102930518A CN 2012101956206 A CN2012101956206 A CN 2012101956206A CN 201210195620 A CN201210195620 A CN 201210195620A CN 102930518 A CN102930518 A CN 102930518A
- Authority
- CN
- China
- Prior art keywords
- image
- resolution
- super
- low
- improved
- 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
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000005070 sampling Methods 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 6
- 239000006185 dispersion Substances 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 230000004304 visual acuity Effects 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 3
- 239000004744 fabric Substances 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 12
- 238000012545 processing Methods 0.000 abstract description 6
- 238000007781 pre-processing Methods 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention aims to disclose an improved sparse representation based image super-resolution method. The improved sparse representation based image super-resolution method comprises the steps of: (1) a training stage; and (2) a reconstruction stage, wherein the training stage comprises sub-steps of: (1) collecting an image sample; (2) extracting image characteristics; (3) sorting the image characteristics; and (4) training a super-complete dictionary; and the reconstruction stage comprises the sub-steps of: (1) pre-processing; (2) extracting characteristics of an image block; (3) coding sparsely; and (4) optimizing globally. Compared with the prior art, as the edge and texture characteristics are adopted, a super-complete dictionary with a better structure is obtained through characteristic classification and sparse coding learning, and the most relevant low-dimensional sub-dictionary can be selected fro an input low-resolution image block for reconstruction. The improved sparse representation based image super-resolution method, disclosed by the invention, has the advantages of improving a processing speed to a certain degree, obtaining a relatively good image effect, realizing the purpose of improving the speed and the effect at the same time, and realizing the purpose of the invention.
Description
Technical field
The present invention relates to a kind of image super-resolution method, particularly a kind of rarefaction representation that adopts is processed the image super-resolution method based on improved rarefaction representation with analysis image.
Background technology
The image super-resolution technology is intended to overcome the restriction of the intrinsic physical features such as resolution of imaging sensor, and low-resolution image is changed into high-definition picture, thereby brings better visual experience.
The mode that improves image resolution ratio has two kinds, and a kind of is the hardware device that directly improves imaging system, and another is the method that adopts signal to process.At present, on physical optics theory, the manufacturing process technology of sensor has reached Limiting Level, can obtain the limiting resolution that hardware provides.In addition, expensive high-precision sensor also orders about us and seeks other solutions.
Development along with image processing techniques and signal processing theory, the low-resolution image of one or more Same Scene has appearred adopting, combining image represents to carry out with technology such as signature analysises the method for image super-resolution, and these methods have the advantages that cost is low, dirigibility is strong.
Image super-resolution is a popular research field of image processing field, propose first to utilize multiframe sequence low-resolution image to come the reconstruct high-definition picture from Huang in 1984 and Tsai, develop into the nowadays advanced image super-resolution method based on rarefaction representation, people are devoted to discussion always both can obtain better effect, can be used for again the method for processing in real time.
But the method for present stage mainly is based on the method for study, and the key of these class methods is feature and the prior model of image representation, directly affects edge and the grain effect of image, and the treatment effeciency of method.
Therefore, need especially a kind of image super-resolution method based on improved rarefaction representation, to solve the above-mentioned existing problem that exists.
Summary of the invention
The object of the present invention is to provide a kind of image super-resolution method based on improved rarefaction representation, for the deficiencies in the prior art, can obtain preferably edge and grain effect, and have good treatment effeciency.
Technical matters solved by the invention can realize by the following technical solutions:
A kind of image super-resolution method based on improved rarefaction representation is characterized in that it comprises the steps:
(1) training stage
1) image pattern collection by a large amount of high-definition picture of camera acquisition, comprises buildings, animal, flowers and plants, people etc.;
2) image characteristics extraction at first, is extracted a pair of image block as characteristics of image at random to height and the low-resolution image sample that collects; Then, according to the experimental bias threshold value that sets characteristics of image is carried out cutting, keep the characteristics of image that contains much information; At last, characteristics of image is carried out normalized;
3) characteristics of image classification is classified according to the characteristics of image that obtains in the above-mentioned steps, each class is had more structural, and preserves all cluster centre values;
4) the super complete dictionary of training according to each category feature that obtains in the above-mentioned steps, adopts the sparse coding training to obtain equating the sub-dictionary of classification, and with above-mentioned steps in the cluster centre value that obtains be merged into as a whole super complete dictionary;
(2) reconstruction stage
1) pre-service is used the bicubic interpolation method to low-resolution image to be measured and is risen the image that sampling obtains target cun size, and image is converted into the YCbCr form;
2) image block characteristics extracts, and extracts the image block characteristics of low-resolution image in the mode of overlapping block, determines according to the cluster centre value that above-mentioned steps obtains which kind of feature is this feature belong to, and also preserves the average pixel value of this image block simultaneously;
3) sparse coding, carry out sparse coding according to the sub-dictionary of low resolution of the corresponding classes of all kinds of feature selecting that obtain in the above-mentioned steps and try to achieve the rarefaction representation coefficient, obtain high-resolution image block characteristics in conjunction with the sub-dictionary of high resolving power, and with above-mentioned steps in the average pixel value addition of the correspondence image piece preserved obtain final high-definition picture piece;
4) global optimization: at first, all high-definition picture pieces of obtaining in the above-mentioned steps are averaged after by the correspondence position stack; Then, with above-mentioned steps in the Cb that obtains and Cr passage merge and be converted into original picture format; At last, utilize iteration back projection method to carry out global optimization after, the Output rusults image.
In one embodiment of the invention, in the step 2 of above-mentioned training stage) in, described low-resolution image is again to obtain through the low-resolution image in the step 1) being carried out the sampling of bicubic interpolation liter.
In one embodiment of the invention, described characteristics of image comprises respectively edge feature and textural characteristics, at first, utilizes the Canny algorithm to extract binary edge map to the high-definition picture piece, asks single order and second order to lead to low-resolution image and obtains four width of cloth gradient images; Then, take the image block central point whether as edge pixel as benchmark, obtain respectively edge image piece and texture image piece.
In one embodiment of the invention, what low-resolution image extracted is Gradient Features, and what high-definition picture extracted is the value that the image block pixel deducts its average pixel value.
In one embodiment of the invention, described experimental bias Threshold is 10, if image block is levied greater than this threshold value, then keeps, otherwise abandons.
In one embodiment of the invention, in the step 3) of above-mentioned training stage, described tagsort process comprises: at first, the low-resolution image feature is classified, and obtain all kinds of cluster centre values; Then, according to the low-resolution image feature of having divided class, in the mode of sitting in the right seat the high-definition picture feature is classified;
In one embodiment of the invention, described cluster centre value has adopted overlapping clustering method, in the situation that keep original cluster centre value constant, make all kinds of marginal points belong to simultaneously a plurality of classes by calculating, reduce the error that causes because of information dispersion.
In one embodiment of the invention, in the step 2 of above-mentioned reconstruction stage) in, the image block size is definite according to the given information of super complete dictionary, and its overlaid pixel value then can freely be determined; Described image block characteristics comprises edge feature and textural characteristics.
In one embodiment of the invention, in the step 4) of above-mentioned reconstruction stage, after average is got in described high-definition picture piece stack, will be that zero pixel is filled to pixel value this moment also, the value of filling be the respective pixel value of the illumination passage of the low-resolution image of liter sampling.
Compared with prior art, the image super-resolution method based on improved rarefaction representation of the present invention has following beneficial effect:
1) edge feature and textural characteristics have been used respectively, than before good based on the method based on rarefaction representation of single features obtained edge effect and grain effect;
2) designed the tagsort stage before the stage at sparse coding, may learn and have more structural dictionary; Also can select the dictionary of maximally related low dimension to be reconstructed for the low-resolution image piece of input in reconstruction stage, therefore can improve treatment effeciency and image effect;
3) propose that overlapping clustering method is arranged, solved the information dispersion problem that causes because of cluster, reduced reconstructed error, further improved image effect.
Image super-resolution method based on improved rarefaction representation of the present invention, compared with prior art adopt edge and texture bicharacteristic, obtain having more structural super complete dictionary by tagsort and sparse coding study, can select the sub-dictionary of maximally related low dimension to be reconstructed for the low-resolution image piece of input, improved to a certain extent processing speed, and obtain preferably image effect, reached the purpose that improves simultaneously speed and effect, realize purpose of the present invention.
Characteristics of the present invention can be consulted the detailed description of the graphic and following better embodiment of this case and be obtained to be well understood to.
Description of drawings
Fig. 1 is the schematic flow sheet of training stage of the present invention;
Fig. 2 is the schematic flow sheet of reconstruction stage of the present invention;
Fig. 3 is the schematic flow sheet of feature extraction of the present invention;
Fig. 4 is the schematic flow sheet that overlapping cluster is arranged of the present invention.
Embodiment
For technological means, creation characteristic that the present invention is realized, reach purpose and effect is easy to understand, below in conjunction with concrete diagram, further set forth the present invention.
As depicted in figs. 1 and 2, the image super-resolution method based on improved rarefaction representation of the present invention, it comprises the steps:
(1) training stage
1) image pattern collection by a large amount of high-definition picture of camera acquisition, comprises buildings, animal, flowers and plants, people etc.;
2) image characteristics extraction at first, is carried a pair of image block as characteristics of image at random to the height and the low-resolution image sample that collect; Then, according to the experimental bias threshold value that sets the image spy is carried out cutting, keep the characteristics of image that contains much information; At last, characteristics of image is carried out normalized;
3) characteristics of image classification is classified according to the characteristics of image that obtains in the above-mentioned steps, each class is had more structural, and preserves all cluster centre values;
4) the super complete dictionary of training according to each category feature that obtains in the above-mentioned steps, adopts the sparse coding training to obtain equating the sub-dictionary of classification, and with above-mentioned steps in the cluster centre value that obtains be merged into as a whole super complete dictionary;
In training dictionary process, setting the degree of rarefication parameter is 0.15, after alternate cycles 40 suboptimization, can obtain thinking the super complete dictionary of convergence.Described dictionary comprises respectively edge dictionary and texture dictionary, and wherein each dictionary comprises high resolving power dictionary and low resolution dictionary;
(2) reconstruction stage
1) pre-service is used the bicubic interpolation method to low-resolution image to be measured and is risen the image that sampling obtains the target size size, and image is converted into the YCbCr form;
2) image block characteristics extracts, and extracts the image block characteristics of low-resolution image in the mode of overlapping block, determines according to the cluster centre value that above-mentioned steps obtains which kind of feature is this feature belong to, and also preserves the average pixel value of this image block simultaneously;
3) sparse coding, carry out sparse coding according to the sub-dictionary of low resolution of the corresponding classes of all kinds of feature selecting that obtain in the above-mentioned steps and try to achieve the rarefaction representation coefficient, obtain high-resolution image block characteristics in conjunction with the sub-dictionary of high resolving power, and with above-mentioned steps in the average pixel value addition of the correspondence image piece preserved obtain final high-definition picture piece;
4) global optimization: at first, all high-definition picture pieces of obtaining in the above-mentioned steps are averaged after by the correspondence position stack; Then, with above-mentioned steps in the Cb that obtains and Cr passage merge and be converted into original picture format; At last, utilize iteration back projection method to carry out global optimization after, output fruit image.
In the present invention, in the step 2 of above-mentioned training stage) in, described low-resolution image is again to obtain through the low-resolution image in the step 1) being carried out the sampling of bicubic interpolation liter.
As shown in Figure 3, described characteristics of image comprises respectively edge feature and textural characteristics, at first, utilizes the Canny algorithm to extract binary edge map to the high-definition picture piece, asks single order and second order to lead to low-resolution image and obtains four width of cloth gradient images; Then, take the image block central point whether as edge pixel as benchmark, obtain respectively edge image piece and texture image piece.
In the present invention, what low-resolution image extracted is Gradient Features, and what high-definition picture extracted is the value that the image block pixel deducts its average pixel value.In fact the tile size that extracts should be 7*7, obtains so at last the low resolution proper vector of 4* (7*7)=196 dimension, the high-resolution features vector of 7*7=49 dimension.
Described experimental bias Threshold is 10, if image block characteristics greater than this threshold value, then keeps, otherwise abandons.
In the present invention, in the step 3) of above-mentioned training stage, described tagsort process comprises: at first, the low-resolution image feature is classified, and obtain all kinds of cluster centre values; Then, according to the low-resolution image feature of having divided class, in the mode of sitting in the right seat the high-definition picture feature is classified.
As shown in Figure 4, described cluster centre value has adopted overlapping clustering method, in the situation that keep original cluster centre value constant, makes all kinds of marginal points belong to simultaneously a plurality of classes by calculating, reduces the error that causes because of information dispersion.Among the present invention, get R=0.1 comparatively suitable.
In the present invention, in the step 2 of above-mentioned reconstruction stage) in, the image block size is definite according to the given information of super complete dictionary, and its overlaid pixel value then can freely be determined; Described image block characteristics comprises edge feature and textural characteristics.
In the present invention, in the step 4) of above-mentioned reconstruction stage, after described high-definition picture piece is folded and got average, to be that zero pixel is filled to pixel value this moment also, the value of filling is the respective pixel value of the illumination passage of the low-resolution image of liter sampling.
Above demonstration and described ultimate principle of the present invention and principal character and advantage of the present invention.The technician of the industry should understand; the present invention is not restricted to the described embodiments; that describes in above-described embodiment and the instructions just illustrates principle of the present invention; without departing from the spirit and scope of the present invention; the present invention also has various changes and modifications; these changes and improvements all fall in the claimed scope of the invention, and the claimed scope of the present invention is defined by appending claims and equivalent thereof.
Claims (9)
1. the image super-resolution method based on improved rarefaction representation is characterized in that, it comprises the steps:
(1) training stage
1) image pattern collection by a large amount of high-definition picture of camera acquisition, comprises buildings, animal, flowers and plants, people etc.;
2) image characteristics extraction at first, is extracted a pair of image block as characteristics of image at random to height and the low-resolution image sample that collects; Then, according to the experimental bias threshold value that sets characteristics of image is carried out cutting, keep the characteristics of image that contains much information; At last, characteristics of image is carried out normalized;
3) characteristics of image classification is classified according to the characteristics of image that obtains in the above-mentioned steps, each class is had more structural, and preserves all cluster centre values;
4) the super complete dictionary of training according to each category feature that obtains in the above-mentioned steps, adopts the sparse coding training to obtain equating the sub-dictionary of classification, and with above-mentioned steps in the cluster centre value that obtains be merged into as a whole super complete dictionary;
(2) reconstruction stage
1) pre-service is used the bicubic interpolation method to low-resolution image to be measured and is risen the image that sampling obtains the target size size, and image is converted into the YCbCr form;
2) image block characteristics extracts, and extracts the image block characteristics of low-resolution image in the mode of overlapping block, determines according to the cluster centre value that above-mentioned steps obtains which kind of feature is this feature belong to, and also preserves the average pixel value of this image block simultaneously;
3) sparse coding, carry out sparse coding according to the sub-dictionary of low resolution of the corresponding classes of all kinds of feature selecting that obtain in the above-mentioned steps and try to achieve the rarefaction representation coefficient, obtain high-resolution image block characteristics in conjunction with the sub-dictionary of high resolving power, and with above-mentioned steps in the average pixel value of the correspondence image piece preserved be added to final high-definition picture piece;
4) global optimization: at first, all high-definition picture pieces of obtaining in the above-mentioned steps are averaged after by the correspondence position stack; Then, with above-mentioned steps in the Cb that obtains and Cr passage merge and be converted into original picture format; At last, utilize iteration back projection method to carry out global optimization after, the Output rusults image.
2. the image super-resolution method based on improved rarefaction representation as claimed in claim 1, it is characterized in that, step 2 in the above-mentioned training stage) in, described low-resolution image is again to obtain through the low-resolution image in the step 1) being carried out the sampling of bicubic interpolation liter.
3. the image super-resolution method based on improved rarefaction representation as claimed in claim 1, it is characterized in that, described characteristics of image comprises respectively edge feature and textural characteristics, at first, utilize the Canny algorithm to extract binary edge map to the high-definition picture piece, ask single order and second order to lead to low-resolution image and obtain four width of cloth gradient images; Then, take the image block central point whether as edge pixel as benchmark, obtain respectively edge image piece and texture image piece.
4. the image super-resolution method based on improved rarefaction representation as claimed in claim 1 is characterized in that, what low-resolution image extracted is Gradient Features, and what high-definition picture extracted is the value that the image block pixel deducts its average pixel value.
5. the image super-resolution method based on improved rarefaction representation as claimed in claim 1 is characterized in that, described experimental bias Threshold is 10, if image block characteristics greater than this threshold value, then keeps, otherwise abandons.
6. the image super-resolution method based on improved rarefaction representation as claimed in claim 1, it is characterized in that, in the step 3) of above-mentioned training stage, described tagsort process comprises: at first, the low-resolution image feature is classified, and obtain all kinds of cluster centre values; Then, according to the low-resolution image feature of having divided class, in the mode of sitting in the right seat high-definition picture feature row is classified.
7. the image super-resolution method based on improved rarefaction representation as claimed in claim 1, it is characterized in that, described cluster centre value has adopted overlapping clustering method, in the situation that keep original cluster centre value constant, make all kinds of marginal points belong to simultaneously a plurality of classes by calculating, reduce the error that causes because of information dispersion.
8. the image super-resolution method based on improved rarefaction representation as claimed in claim 1, it is characterized in that, step 2 in above-mentioned reconstruction stage) in, the image block size is definite according to the given information of super complete dictionary, and its overlaid pixel value then can freely be determined; Described image block characteristics comprises edge feature and textural characteristics.
9. the image super-resolution method based on improved rarefaction representation as claimed in claim 1, it is characterized in that, in the step 4) of above-mentioned reconstruction stage, after average is got in described high-definition picture piece stack, will be that zero pixel is filled to pixel value this moment also, the value of filling be the respective pixel value of the illumination passage of the low-resolution image of liter sampling.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210195620.6A CN102930518B (en) | 2012-06-13 | 2012-06-13 | Improved sparse representation based image super-resolution method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210195620.6A CN102930518B (en) | 2012-06-13 | 2012-06-13 | Improved sparse representation based image super-resolution method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102930518A true CN102930518A (en) | 2013-02-13 |
| CN102930518B CN102930518B (en) | 2015-06-24 |
Family
ID=47645308
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210195620.6A Active CN102930518B (en) | 2012-06-13 | 2012-06-13 | Improved sparse representation based image super-resolution method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102930518B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104113789A (en) * | 2014-07-10 | 2014-10-22 | 杭州电子科技大学 | On-line video abstraction generation method based on depth learning |
| CN104677309A (en) * | 2015-03-24 | 2015-06-03 | 哈尔滨理工大学 | Highlight inhibition method for strong reflection surface encoding light measurement based on sparse expression |
| WO2015180055A1 (en) * | 2014-05-28 | 2015-12-03 | 北京大学深圳研究生院 | Super-resolution image reconstruction method and apparatus based on classified dictionary database |
| CN105678728A (en) * | 2016-01-19 | 2016-06-15 | 西安电子科技大学 | High-efficiency super-resolution imaging device and method with regional management |
| CN106600536A (en) * | 2016-12-14 | 2017-04-26 | 同观科技（深圳）有限公司 | Video imager super-resolution reconstruction method and apparatus |
| CN107967669A (en) * | 2017-11-24 | 2018-04-27 | 腾讯科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of picture processing |
| WO2018120329A1 (en) * | 2016-12-28 | 2018-07-05 | 深圳市华星光电技术有限公司 | Single-frame super-resolution reconstruction method and device based on sparse domain reconstruction |
| CN108830792A (en) * | 2018-05-09 | 2018-11-16 | 浙江师范大学 | A kind of image super-resolution method using multiclass dictionary |
| WO2018214671A1 (en) * | 2017-05-26 | 2018-11-29 | 杭州海康威视数字技术股份有限公司 | Image distortion correction method and device and electronic device |
| US10417996B2 (en) | 2017-08-31 | 2019-09-17 | Yuan Ze University | Method, image processing device, and display system for power-constrained image enhancement |
| CN111612691A (en) * | 2020-04-17 | 2020-09-01 | 重庆大学 | Image super-resolution processing improvement method based on sparse representation |
| CN114697543A (en) * | 2020-12-31 | 2022-07-01 | 华为技术有限公司 | Image reconstruction method, related device and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101556690A (en) * | 2009-05-14 | 2009-10-14 | 复旦大学 | Image super-resolution method based on overcomplete dictionary learning and sparse representation |
| CN102142137A (en) * | 2011-03-10 | 2011-08-03 | 西安电子科技大学 | High-resolution dictionary based sparse representation image super-resolution reconstruction method |
| CN102156875A (en) * | 2011-03-25 | 2011-08-17 | 西安电子科技大学 | Image super-resolution reconstruction method based on multitask KSVD (K singular value decomposition) dictionary learning |
-
2012
- 2012-06-13 CN CN201210195620.6A patent/CN102930518B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101556690A (en) * | 2009-05-14 | 2009-10-14 | 复旦大学 | Image super-resolution method based on overcomplete dictionary learning and sparse representation |
| CN102142137A (en) * | 2011-03-10 | 2011-08-03 | 西安电子科技大学 | High-resolution dictionary based sparse representation image super-resolution reconstruction method |
| CN102156875A (en) * | 2011-03-25 | 2011-08-17 | 西安电子科技大学 | Image super-resolution reconstruction method based on multitask KSVD (K singular value decomposition) dictionary learning |
Non-Patent Citations (3)
| Title |
|---|
| 李民等: "稀疏字典编码的超分辨率重建", 《软件学报》 * |
| 浦剑等: "基于词典学习和稀疏表示的超分辨率方法", 《模式识别与人工智能》 * |
| 练秋生等: "基于图像块分类稀疏表示的超分辨率重构算法", 《电子学报》 * |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015180055A1 (en) * | 2014-05-28 | 2015-12-03 | 北京大学深圳研究生院 | Super-resolution image reconstruction method and apparatus based on classified dictionary database |
| CN104113789A (en) * | 2014-07-10 | 2014-10-22 | 杭州电子科技大学 | On-line video abstraction generation method based on depth learning |
| CN104113789B (en) * | 2014-07-10 | 2017-04-12 | 杭州电子科技大学 | On-line video abstraction generation method based on depth learning |
| CN104677309A (en) * | 2015-03-24 | 2015-06-03 | 哈尔滨理工大学 | Highlight inhibition method for strong reflection surface encoding light measurement based on sparse expression |
| CN104677309B (en) * | 2015-03-24 | 2017-04-26 | 哈尔滨理工大学 | Highlight inhibition method for strong reflection surface encoding light measurement based on sparse expression |
| CN105678728A (en) * | 2016-01-19 | 2016-06-15 | 西安电子科技大学 | High-efficiency super-resolution imaging device and method with regional management |
| CN106600536A (en) * | 2016-12-14 | 2017-04-26 | 同观科技（深圳）有限公司 | Video imager super-resolution reconstruction method and apparatus |
| CN106600536B (en) * | 2016-12-14 | 2020-02-14 | 同观科技（深圳）有限公司 | Video image super-resolution reconstruction method and device |
| WO2018120329A1 (en) * | 2016-12-28 | 2018-07-05 | 深圳市华星光电技术有限公司 | Single-frame super-resolution reconstruction method and device based on sparse domain reconstruction |
| WO2018214671A1 (en) * | 2017-05-26 | 2018-11-29 | 杭州海康威视数字技术股份有限公司 | Image distortion correction method and device and electronic device |
| CN108932697A (en) * | 2017-05-26 | 2018-12-04 | 杭州海康威视数字技术股份有限公司 | A kind of distorted image removes distortion methods, device and electronic equipment |
| CN108932697B (en) * | 2017-05-26 | 2020-01-17 | 杭州海康威视数字技术股份有限公司 | Distortion removing method and device for distorted image and electronic equipment |
| US11250546B2 (en) | 2017-05-26 | 2022-02-15 | Hangzhou Hikvision Digital Technology Co., Ltd. | Image distortion correction method and device and electronic device |
| US10417996B2 (en) | 2017-08-31 | 2019-09-17 | Yuan Ze University | Method, image processing device, and display system for power-constrained image enhancement |
| CN107967669A (en) * | 2017-11-24 | 2018-04-27 | 腾讯科技（深圳）有限公司 | Method, apparatus, computer equipment and the storage medium of picture processing |
| CN108830792A (en) * | 2018-05-09 | 2018-11-16 | 浙江师范大学 | A kind of image super-resolution method using multiclass dictionary |
| CN108830792B (en) * | 2018-05-09 | 2022-03-11 | 浙江师范大学 | Image super-resolution method using multi-class dictionary |
| CN111612691A (en) * | 2020-04-17 | 2020-09-01 | 重庆大学 | Image super-resolution processing improvement method based on sparse representation |
| CN114697543A (en) * | 2020-12-31 | 2022-07-01 | 华为技术有限公司 | Image reconstruction method, related device and system |
| WO2022143921A1 (en) * | 2020-12-31 | 2022-07-07 | 华为技术有限公司 | Image reconstruction method, and related apparatus and system |
| CN114697543B (en) * | 2020-12-31 | 2023-05-19 | 华为技术有限公司 | Image reconstruction method, related device and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102930518B (en) | 2015-06-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102930518B (en) | Improved sparse representation based image super-resolution method | |
| WO2021184891A1 (en) | Remotely-sensed image-based terrain classification method, and system | |
| CN112287940B (en) | Semantic segmentation method of attention mechanism based on deep learning | |
| CN109671023B (en) | Face image super-resolution secondary reconstruction method | |
| Li et al. | Joint example-based depth map super-resolution | |
| CN101976435B (en) | Combination learning super-resolution method based on dual constraint | |
| CN105046672A (en) | Method for image super-resolution reconstruction | |
| Hua et al. | A normalized convolutional neural network for guided sparse depth upsampling. | |
| CN105069746A (en) | Video real-time human face substitution method and system based on partial affine and color transfer technology | |
| CN105335929A (en) | Depth map super-resolution method | |
| CN104751432A (en) | Image reconstruction based visible light and infrared image fusion method | |
| CN104899835B (en) | Image Super-resolution processing method based on blind blur estimation and anchoring space mapping | |
| CN103984963B (en) | Method for classifying high-resolution remote sensing image scenes | |
| CN103854268A (en) | Image super-resolution reconstruction method based on multi-core gaussian process regression | |
| CN104036468A (en) | Super-resolution reconstruction method for single-frame images on basis of pre-amplification non-negative neighbor embedding | |
| KR101028628B1 (en) | Image texture filtering method, storage medium of storing program for executing the same and apparatus performing the same | |
| CN103714526A (en) | Super-resolution image reconstruction method based on sparse multi-manifold embedment | |
| CN112560701B (en) | Face image extraction method and device and computer storage medium | |
| CN113610024B (en) | Multi-strategy deep learning remote sensing image small target detection method | |
| CN108230223A (en) | Light field angle super-resolution rate method and device based on convolutional neural networks | |
| CN110111272A (en) | A kind of artificial intelligence infrared image processing instrument, control system and control method | |
| CN108460833A (en) | A kind of information platform building traditional architecture digital protection and reparation based on BIM | |
| CN104021567B (en) | Based on the fuzzy altering detecting method of image Gauss of first numeral law | |
| CN103971354A (en) | Method for reconstructing low-resolution infrared image into high-resolution infrared image | |
| CN112766123A (en) | Crowd counting method and system based on criss-cross attention network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C53 | Correction of patent of invention or patent application | ||
| CB02 | Change of applicant information |
Address after: 201505 Jinshan District, Shanghai, Lin Ting Town, Temple Ping Road, room 16, No. 599 Applicant after: SHANGHAI WINNER INFORMATION TECHNOLOGY CO., INC. Address before: 201505 Jinshan District, Shanghai, Lin Ting Town, Temple Ping Road, room 16, No. 599 Applicant before: Huina Network Information Science & Technology Co., Ltd. |
|
| COR | Change of bibliographic data |
Free format text: CORRECT: APPLICANT; FROM: HUINA NETWORK INFORMATION SCIENCE + TECHNOLOGY CO., LTD. TO: SHANGHAI WINNER INFORMATION TECHNOLOGY CO., INC. |
|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 201505 Shanghai city Jinshan District ting Lin Zhen Ting Feng Road No. 333, room 216 Patentee after: Winner Technology Co.,Ltd. Address before: 201505, room 16, No. 599, Ping Nan Road, ting Lin Town, Shanghai, Jinshan District Patentee before: SHANGHAI WINNER INFORMATION TECHNOLOGY Co.,Inc. |
|
| CP03 | Change of name, title or address | ||
| CP02 | Change in the address of a patent holder | ||
| CP02 | Change in the address of a patent holder |
Address after: 201203 No. 6, Lane 55, Chuanhe Road, China (Shanghai) pilot Free Trade Zone, Pudong New Area, Shanghai Patentee after: Winner Technology Co.,Ltd. Address before: 201505 Room 216, 333 Tingfeng Highway, Tinglin Town, Jinshan District, Shanghai Patentee before: Winner Technology Co.,Ltd. |