CN108537734A - Single image super resolution ratio reconstruction method based on gradient profile example dictionary and Weighted adaptive p norms - Google Patents
Single image super resolution ratio reconstruction method based on gradient profile example dictionary and Weighted adaptive p norms Download PDFInfo
- Publication number
- CN108537734A CN108537734A CN201810335606.9A CN201810335606A CN108537734A CN 108537734 A CN108537734 A CN 108537734A CN 201810335606 A CN201810335606 A CN 201810335606A CN 108537734 A CN108537734 A CN 108537734A
- Authority
- CN
- China
- Prior art keywords
- resolution
- image
- adaptive
- norm
- gradient profile
- 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 120
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 35
- 238000003780 insertion Methods 0.000 claims abstract description 5
- 230000037431 insertion Effects 0.000 claims abstract description 5
- 238000005457 optimization Methods 0.000 claims abstract description 4
- 239000000284 extract Substances 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 2
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 claims 2
- 239000011159 matrix material Substances 0.000 claims 1
- 238000011156 evaluation Methods 0.000 abstract description 7
- 238000002474 experimental method Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000001419 dependent effect Effects 0.000 description 2
- 238000011017 operating method Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4007—Interpolation-based scaling, e.g. bilinear interpolation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4053—Super resolution, i.e. output image resolution higher than sensor resolution
-
- G06T5/70—
-
- G06T5/73—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Processing (AREA)
Abstract
Denomination of invention is based on gradient profile example dictionary and Weighted adaptivepThe invention discloses one kind being based on gradient profile example dictionary and Weighted adaptive for the single image super resolution ratio reconstruction method abstract of normpThe single image super resolution ratio reconstruction method of norm.It mainly includes the following steps that：The low-resolution image of input is up-sampled using interpolation method；The estimation of high-resolution gradient image is completed by the neighborhood insertion super resolution ratio reconstruction method based on gradient profile example dictionary；Build the similar block group of form adaptive；Establish the weighting of form adaptive similar block group singular valuepNorm constraint, and each similar block group norm constraint is adaptively adjusted according to the conspicuousness of image-regionpValue；It establishes and rebuilds cost function, and iteration optimization, obtain the high-definition picture of final output.It is of the present invention to be based on gradient profile example dictionary and Weighted adaptivepThe single image super resolution ratio reconstruction method of norm has preferable subjective vision and higher objective evaluation value.Therefore, the present invention is a kind of effective single image super resolution ratio reconstruction method.
Description
Technical field
The present invention relates to image super-resolution rebuilding technologies, and in particular to one kind being based on gradient profile example dictionary and weighting
AdaptivelypThe single image super resolution ratio reconstruction method of norm, belongs to digital image processing field.
Background technology
In actual life, limited by factors such as image system hardware equipment, image-forming condition, information transmission conditions, people
Can not usually obtain high-resolution observed image.Image super-resolution rebuilding technology, can be in the feelings for not increasing hardware cost
The constraint that these restrictive conditions are broken through under condition has the low resolution observed image of complementary information using one or more, rebuilds
A panel height image in different resolution is obtained, allows one to more fully understand picture material and for further analysis and utilization.With several
Image super-resolution technology is compared, and the required low resolution observed image number of single image super-resolution technique is less, makes
It uses more flexible.Correspondingly, from the point of view of signal processing, due to the deficiency of constraint information, single image super-resolution is enabled
Difficulty and challenge of the solution with bigger of Problems of Reconstruction.
It includes mainly three research directions that single image super-resolution rebuilding technology, which is developed so far,：Method based on interpolation,
Method based on reconstruction and the method based on study.Method arithmetic speed based on interpolation is fast, but is readily incorporated fuzzy and saw
Tooth effect.Based on the method for study by learning and using the relationship between low-resolution image and high-definition picture, realizing and divide
The raising of resolution.Method based on study can preferably restoring image detail, but performance is highly dependent on training set and test set
Similarity degree.Method based on reconstruction makes reconstruction image have dependent on degrade model and various image priors, the use of priori
For corresponding natural image statistical property.It will be apparent that better Super-resolution reconstruction can be obtained using more accurate prior model
Build performance.
Invention content
The purpose of the present invention is be embedded in combine with gradient profile feature by neighborhood to complete high-resolution gradient profile
Estimation, and introduce super-resolution rebuilding using the high-resolution gradient image of synthesis as partial gradient region constraint priori；By right
The weighting of non local similar block group singular valuepNorm constraint realizes the low-rank characteristic of block group, adaptive with the conspicuousness of image-region
Each similar block group should be adjustedpValue improvespThe flexibility of norm constraint and accuracy, and by the Weighted adaptive of foundationpModel
Exponential model introduces super-resolution rebuilding；By the joint of two kinds of local and non local priori, the performance of super-resolution rebuilding is improved,
Reconstruction is set effectively to inhibit various distortions, preferably restoring image detail and sharpening edge.The present invention passes through following operating procedure
The technical solution of composition realizes above-mentioned purpose.
It is proposed by the present invention to be based on gradient profile example dictionary and Weighted adaptivepThe single image Super-resolution reconstruction of norm
Construction method includes mainly following operating procedure：
（1）The low-resolution image of input is up-sampled using bicubic interpolation method, initial high-resolution is obtained and estimates
Count image；
（2）High-resolution gradient image is completed by the neighborhood insertion super resolution ratio reconstruction method based on gradient profile example dictionary
Estimation builds gradient field priori；
（3）Estimate image according to existing high-resolution, extracts the adaptive neighborhood of each reference pixel, and carry out non local phase
Like block search, the adaptive similar block group of formed shape；
（4）Estimate image according to existing high-resolution, determines each reference pixelpNorm value, and build form adaptive
The Weighted adaptive of similar block grouppNorm priori；
（5）The Weighted adaptive that gradient field priori, step 4 that the low-resolution image of input and step 2 obtain are obtainedpModel
Number priori is established as constraint and rebuilds cost function；
（6）Step (3) ~ step (5) is repeated, the optimization for rebuilding cost function is completed.When meeting stopping criterion for iteration,
Execute step (7)；
（7）Meet stopping criterion for iteration, estimates image as final high-resolution weight the high-resolution of last time iteration
Build image output.
Description of the drawings
Fig. 1 is that the present invention is based on gradient profile example dictionary and Weighted adaptivespThe single image Super-resolution reconstruction of norm
The block diagram of construction method.
Fig. 2 is that the present invention is based on the block diagrams that the gradient image of gradient profile example dictionary is estimated.
Fig. 3 is comparison diagram of the present invention with existing 10 kinds of methods to image " Foreman " reconstructed results.
Fig. 4 is the present invention and existing 8 kinds of methods to noise image（Noise criteria difference is 10）" Eyetest " rebuilds knot
The comparison diagram of fruit.
Specific implementation mode
The invention will be further described below in conjunction with the accompanying drawings：
In Fig. 1, it is based on gradient profile example dictionary and Weighted adaptivepThe single image super resolution ratio reconstruction method of norm, packet
Include following steps：
（1）It to the low-resolution image of input, is up-sampled using bicubic interpolation method, obtains initial high-resolution and estimate
Count image；
（2）High-resolution gradient image is completed by the neighborhood insertion super resolution ratio reconstruction method based on gradient profile example dictionary
Estimation, gradient field priori is built with the high-resolution gradient image of estimation；
（3）Estimate image according to existing high-resolution, extracts the adaptive neighborhood of each reference pixel, and carry out non local phase
Like block search, the adaptive similar block group of formed shape；
（4）Estimate image according to existing high-resolution, determines each reference pixelpNorm value, and build form adaptive
The Weighted adaptive of similar block grouppNorm priori；
（5）The Weighted adaptive that gradient field priori, step 4 that the low-resolution image of input and step 2 obtain are obtainedpModel
Number priori is established as constraint and rebuilds cost function；
（6）Step (3) ~ step (5) is repeated, the optimization for rebuilding cost function is completed.When meeting stopping criterion for iteration,
Execute step (7)；
（7）Meet stopping criterion for iteration, estimates image as final high-resolution weight the high-resolution of last time iteration
Build image output.
Specifically, the step（1）In, our low-resolution images to input are carried out using bicubic interpolation method
Up-sampling obtains initial high-resolution estimation image.
The step（2）In, we combine neighborhood embedding grammar and image gradient contour feature, to complete high-resolution ladder
Spend the estimation of image.The process includes mainly two stages, i.e. gradient profile example dictionary construction phase and high-resolution gradient
The image reconstruction stage.
In gradient profile example dictionary construction phase, marginal point is detected using Canny operators to training image, is then carried
The low high resolution gradient profile pair centered on these marginal points is taken out, low high resolution gradient profile example word is constituted
Allusion quotation pair.The threshold value of Canny operators should be sufficiently low, so that dictionary includes the gradient profile pair at different strong and weak edges.
The step（3）In, image is estimated to existing high-resolution, we extract often using kernel regression method is oriented to
The shape-adaptive neighborhood of a reference image vegetarian refreshments is usedM i Indicate shape-adaptive neighborhood mask.To giving reference image block, according to
Distance metric criterion searches for non local similar block, forms non local similar block group, and be superimposed with mask to block groupM i Form shape
Adaptive similar block group.
The step（4）In, we are weighted the singular value of each form adaptive similar block grouppNorm constraint,
It is defined as：
Underscore indicates to integrate the Weighted adaptive of each form adaptive similar block group in formulapNorm constraint.
The step（5）In, we are by the low-resolution image and step of input（2）Obtained gradient field priori, step
（4）Obtained Weighted adaptivepNorm priori establishes reconstruction cost function as constraint, rebuilds cost function and is defined as：
The step（6）In, we using division Donald Bragg graceful iterative method (Split Bregman Iteration,
SBI it) optimizes reconstruction function, and compares current iteration indexkWith maximum iterationK_max.Ifk < K_max, then
Repeat step（3）~ step（5）, carry out loop iteration next time；Ifk =K_max, then iteration ends, execute step（7）, willK_maxHigh-resolution in secondary iteration estimates image as final high-definition pictureXOutput.
Validity in order to better illustrate the present invention, the present invention will carry out rebuilding effect using the method for contrast experiment
Displaying.
It is test image, such as Fig. 3 that experiment 1 and experiment 2 have chosen " Foreman " and " Eyetest " respectively（a）And Fig. 4
（a）It is shown.Contrast experiment chooses bicubic interpolation Bicubic and 9 representative single image super-resolution rebuilding sides
Method is compared with the experimental result of the present invention.This 9 representative single image super resolution ratio reconstruction methods are：
Method 1：The method that Yang et al. is proposed, bibliography " J. Yang, J. Wright, T. S. Huang, and
Y. Ma, “Image super-resolution via sparse representation,” IEEE Trans. Image Process., vol. 19, no. 11, pp. 2861-2873, Nov. 2010.”。
Method 2：The method that Dong et al. is proposed, bibliography " C. Dong, C. C. Loy, K. He, and X.
Tang, “Image super-resolution using deep convolutional networks,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 38, no. 2, pp. 295-307, Feb. 2016.”。
Method 3：The method that Zhang et al. is proposed, bibliography " K. Zhang, X. Gao, D. Tao, and X.
Li, “Single image super-resolution with non-local means and steering kernel
regression,” IEEE Trans. Image Process., vol. 21, no. 11, pp. 4544-4556, Nov.
2012.”。
Method 4：The method that Timofte et al. is proposed, bibliography " R. Timofte, V. De Smet, and L.
Van Gool, “A+: Adjusted anchored neighborhood regression for fast super-
resolution,” in Proc. Asian Conf. Comput. Vis. (ACCV), 2014, pp. 111-126.”。
Method 5：The method that Tian et al. is proposed, bibliography " Y. Tian, F. Zhou, W. Yang, X.
Shang, and Q. Liao, “Anchored neighborhood regression based single image
super-resolution from self-examples,” in Proc. IEEE Int. Conf. Image Process (ICIP), Sep. 2016, pp. 2827-2831.”。
Method 6：The method that Dong et al. is proposed, bibliography " W. Dong, L. Zhang, G. Shi, and X.
Wu, “Image deblurring and superresolution by adaptive sparse domain selection
and adaptive regularization,” IEEE Trans. Image Process., vol. 20, no. 7, pp.
1838-1857, Jul. 2011.”。
Method 7：The method that Dong et al. is proposed, bibliography " W. Dong, L. Zhang, G. Shi, and X.
Li, “Nonlocally centralized sparse representation for image restoration,”IEEE Trans. Image Process., vol. 22, no. 4, pp. 1620-1630, Apr. 2013.”。
Method 8：The method that Papyan et al. is proposed, bibliography " V. Papyan and M. Elad, " Multi-
scale patch-based image restoration,” IEEE Trans. Image Process., vol. 25,
no. 1, pp. 249-261, Jan. 2016.”。
Method 9：The method that Chen et al. is proposed, bibliography " H. Chen, X. He, L. Qing, and Q.
Teng, “Single Image Super-Resolution via Adaptive Transform-Based Nonlocal
Self-Similarity Modeling and Learning-Based Gradient Regularization,” IEEE Trans. Multimedia, vol. 19, no. 8, pp. 1702-1717, Aug. 2017.”。
The content of contrast experiment is as follows：
Bicubic, method 1, method 2, method 3, method 4, method 5, method 6, method 7, method 8, method 9 are used in experiment 1 respectively
And the present invention carries out 3 times of super-resolution rebuildings to image " Foreman ".Reconstructed results are respectively such as Fig. 3（b）, Fig. 3（c）, Fig. 3
（d）, Fig. 3（e）, Fig. 3（f）, Fig. 3（g）, Fig. 3（h）, Fig. 3（i）, Fig. 3（j）, Fig. 3（k）And Fig. 3（l）It is shown, objective evaluation
Index is as shown in Table 1.
Bicubic, method 1, method 2, method 4, method 6, method 7, method 8, method 9 and this hair are used in experiment 2 respectively
It is bright to noise image " Eyetest "（Noise criteria difference is 10）Carry out 3 times of super-resolution rebuildings.Reconstructed results are respectively such as Fig. 4
（b）, Fig. 4（c）, Fig. 4（d）, Fig. 4（e）, Fig. 4（f）, Fig. 4（g）, Fig. 4（h）, Fig. 4（i）And Fig. 4（j）It is shown, objective evaluation
Index is as shown in Table 2.
By experiment 1 as can be seen that the obtained reconstruction images of Biucbic are very fuzzy and have serious sawtooth；1 He of method
Edge is by excess smoothness in 5 result figure of method；Although 4 energy recovered part image detail of method 2 and method, subjective effect are too late
Method 3 based on reconstruction and method 8.But there are still serious reconstruction distortions in 8 reconstructed results of method 3 and method（Especially side
Edge region）.Method 6, method 7 and method 9 have preferably restored image detail, and maintain edge sharpening degree, but attached to edge
Close ring distortion suppression is insufficient.The reconstruction image that the present invention obtains not only effectively inhibits various distortions, also better restores
Image detail and edge is sharpened, there is preferable visual effect.
By experiment 2 as can be seen that the method for being mostly based on study（Such as method 1 and method 4）Have in result figure tight
The noise residual of weight.Although method 2 can inhibit noise to a certain extent, distortion is rebuild there are still apparent.Method 6, side
Method 7, method 8 and method 9 can effectively inhibit noise, but still not as good as this method in terms of details and sharpening Edge restoration.This method
Noise, preferable restoring image detail can effectively be inhibited and sharpen edge, present preferable subjective vision effect.
Table one gives the objective comparison of the present invention and 10 kinds of single image super-resolution rebuilding control methods reconstructed results
Value, table two give the present invention with 8 kinds of single image super-resolution rebuilding control methods to the objective of noise image reconstructed results
Fiducial value.Objective fiducial value has used Y-PSNR respectively（PSNR：the Peak Signal to Noise Ratio）, knot
Structure similarity（SSIM：the Structure Similarity Index）And fidelity of information（IFC： the
information fidelity criterion）These three objective evaluation indexs.Wherein, PSNR values are bigger, SSIM values more connect
It is bordering on that 1, IFC values are bigger, then the quality of reconstruction image is better.
Table one
| Objective indicator | Bicubic | Method 1 | Method 2 | Method 3 | Method 4 | Method 5 |
| PSNR (dB) | 30.74 | 34.45 | 35.21 | 35.74 | 35.56 | 35.09 |
| SSIM | 0.8893 | 0.9317 | 0.9350 | 0.9397 | 0.9391 | 0.9364 |
| IFC | 2.58 | 3.72 | 3.74 | 3.97 | 3.95 | 3.92 |
| Objective indicator | Method 6 | Method 7 | Method 8 | Method 9 | The present invention | |
| PSNR (dB) | 35.58 | 35.97 | 35.26 | 36.20 | 36.72 | |
| SSIM | 0.9400 | 0.9418 | 0.9354 | 0.9423 | 0.9476 | |
| IFC | 4.04 | 4.14 | 3.92 | 4.06 | 4.30 |
Table two
| Objective indicator | Bicubic | Method 1 | Method 2 | Method 3 | Method 6 | Method 7 | Method 8 | Method 9 | The present invention |
| PSNR (dB) | 17.76 | 19.54 | 20.48 | 20.31 | 20.07 | 21.03 | 21.22 | 21.32 | 22.16 |
| SSIM | 0.7278 | 0.7957 | 0.8435 | 0.7542 | 0.8280 | 0.8706 | 0.8837 | 0.8753 | 0.9000 |
| IFC | 2.00 | 2.81 | 3.04 | 3.08 | 3.10 | 3.52 | 3.58 | 3.67 | 3.87 |
As can be seen from Table I, the present invention has higher objective evaluation index.Compared with the more outstanding method 9 of performance, this hair
Bright PSNR values have been higher by 0.52dB, and SSIM values have been higher by 0.0053, IFC values and have been higher by 0.24.From table 2 it can be seen that for
The super-resolution rebuilding of noise image, the present invention equally have higher objective evaluation index.It is of the invention compared with method 9
PSNR values have been higher by 0.84dB, and SSIM values have been higher by 0.0247, IFC values and have been higher by 0.2.
In conclusion the reconstructed results of the present invention have preferable subjective vision and higher objective evaluation value, it is a kind of
Effective single image super resolution ratio reconstruction method.
Claims (4)
1. being based on gradient profile example dictionary and Weighted adaptivepThe single image super resolution ratio reconstruction method of norm, feature
Include the following steps：
Step 1：The low-resolution image of input is up-sampled using bicubic interpolation method, obtains initial high-resolution
Rate estimates image；
Step 2：High-resolution gradient map is completed by the neighborhood insertion super resolution ratio reconstruction method based on gradient profile example dictionary
The estimation of picture builds gradient field priori；
Step 3：Estimate image according to existing high-resolution, extracts the adaptive neighborhood of each reference pixel, and carry out non-office
The similar block search in portion, the adaptive similar block group of formed shape；
Step 4：Estimate image according to existing high-resolution, determines each reference pixelpNorm value, and build shape certainly
Adapt to the Weighted adaptive of similar block grouppNorm priori；
Step 5：The weighting that gradient field priori, step 4 that the low-resolution image of input and step 2 obtain are obtained is adaptive
It answerspNorm priori is established as constraint and rebuilds cost function；
Step 6：Step 3 ~ step 5 is repeated, the optimization for rebuilding cost function is completed, when meeting stopping criterion for iteration
When, execute step 7；
Step 7：Meet stopping criterion for iteration, estimates image as final high-resolution the high-resolution of last time iteration
Rate reconstruction image exports.
2. according to claim 1 be based on gradient profile example dictionary and Weighted adaptivepThe single image oversubscription of norm
Resolution method for reconstructing, it is characterised in that the neighborhood based on gradient profile example dictionary described in step 2 is embedded in super-resolution rebuilding
Method：In conjunction with neighborhood insertion and gradient profile feature, gradient profile example dictionary is established；Low point is extracted from low-resolution image
Resolution gradient profile is searched in low resolution gradient profile example dictionaryKA arest neighbors gives low resolution according to minimizing
Gradient profile andKThe distance between the weighted array of a arest neighbors low resolution gradient profile, determines combining weights, byKIt is a most
The weighted array of the corresponding high-resolution gradient profile of neighbour obtains high-resolution gradient profile to be estimated, finally by all
High-resolution gradient profile synthesizes a panel height resolution gradient image.
3. according to claim 1 be based on gradient profile example dictionary and Weighted adaptivepThe single image oversubscription of norm
Resolution method for reconstructing, it is characterised in that the Weighted adaptive of the form adaptive similar block group described in step 4pNorm priori mould
Type：In order to make form adaptive similar block group meet low-rank characteristic, we establish the weighting of block group singular valuepNorm constraint；
In order to improvepThe flexibility of norm constraint and accuracy, each similar block group are correspondingpNorm value will be according to the aobvious of image-region
Work property adaptively adjusts, and establishes the Weighted adaptive of form adaptive similar block grouppNorm priori.
4. according to claim 1 be based on gradient profile example dictionary and Weighted adaptivepThe single image oversubscription of norm
Resolution method for reconstructing, it is characterised in that the gradient field elder generation that the low-resolution image by input described in step 5 is obtained with step 2
It tests, the Weighted adaptive that step 4 obtainspNorm priori establishes reconstruction cost function as constraint, rebuilds cost function definition
For：
In formula,XFor high-definition picture to be reconstructed,YFor the low-resolution image of input,DDegrade for down-sampling,HIt is fuzzy
Degrade,For step 2 estimation high-resolution gradient image,MFor shape-adaptive neighborhood mask,PIt is adaptivepNorm
Matrix,For the Weighted adaptive for the form adaptive similar block group that step 4 obtainspNorm priori.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810335606.9A CN108537734A (en) | 2018-04-16 | 2018-04-16 | Single image super resolution ratio reconstruction method based on gradient profile example dictionary and Weighted adaptive p norms |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810335606.9A CN108537734A (en) | 2018-04-16 | 2018-04-16 | Single image super resolution ratio reconstruction method based on gradient profile example dictionary and Weighted adaptive p norms |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108537734A true CN108537734A (en) | 2018-09-14 |
Family
ID=63480166
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810335606.9A Pending CN108537734A (en) | 2018-04-16 | 2018-04-16 | Single image super resolution ratio reconstruction method based on gradient profile example dictionary and Weighted adaptive p norms |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108537734A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109741263A (en) * | 2019-01-11 | 2019-05-10 | 四川大学 | Remote sensed image super-resolution reconstruction algorithm based on adaptive combined constraint |
| CN110020989A (en) * | 2019-05-23 | 2019-07-16 | 西华大学 | A kind of depth image super resolution ratio reconstruction method based on deep learning |
| CN112150354A (en) * | 2019-06-26 | 2020-12-29 | 四川大学 | Single image super-resolution method combining contour enhancement and denoising statistical prior |
| CN112862688A (en) * | 2021-03-08 | 2021-05-28 | 西华大学 | Cross-scale attention network-based image super-resolution reconstruction model and method |
| CN115115654A (en) * | 2022-06-14 | 2022-09-27 | 北京空间飞行器总体设计部 | Object image segmentation method based on saliency and neighbor shape query |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104036468A (en) * | 2014-06-19 | 2014-09-10 | 西安电子科技大学 | Super-resolution reconstruction method for single-frame images on basis of pre-amplification non-negative neighbor embedding |
| US8938118B1 (en) * | 2012-12-12 | 2015-01-20 | Rajiv Jain | Method of neighbor embedding for OCR enhancement |
| CN105550988A (en) * | 2015-12-07 | 2016-05-04 | 天津大学 | Super-resolution reconstruction algorithm based on improved neighborhood embedding and structure self-similarity |
| CN105844590A (en) * | 2016-03-23 | 2016-08-10 | 武汉理工大学 | Image super-resolution reconstruction method and system based on sparse representation |
| CN107067367A (en) * | 2016-09-08 | 2017-08-18 | 南京工程学院 | A kind of Image Super-resolution Reconstruction processing method |
| CN107085826A (en) * | 2017-04-11 | 2017-08-22 | 四川大学 | Based on the non local single image super resolution ratio reconstruction method for returning priori of weighted overlap-add |
-
2018
- 2018-04-16 CN CN201810335606.9A patent/CN108537734A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8938118B1 (en) * | 2012-12-12 | 2015-01-20 | Rajiv Jain | Method of neighbor embedding for OCR enhancement |
| CN104036468A (en) * | 2014-06-19 | 2014-09-10 | 西安电子科技大学 | Super-resolution reconstruction method for single-frame images on basis of pre-amplification non-negative neighbor embedding |
| CN105550988A (en) * | 2015-12-07 | 2016-05-04 | 天津大学 | Super-resolution reconstruction algorithm based on improved neighborhood embedding and structure self-similarity |
| CN105844590A (en) * | 2016-03-23 | 2016-08-10 | 武汉理工大学 | Image super-resolution reconstruction method and system based on sparse representation |
| CN107067367A (en) * | 2016-09-08 | 2017-08-18 | 南京工程学院 | A kind of Image Super-resolution Reconstruction processing method |
| CN107085826A (en) * | 2017-04-11 | 2017-08-22 | 四川大学 | Based on the non local single image super resolution ratio reconstruction method for returning priori of weighted overlap-add |
Non-Patent Citations (3)
| Title |
|---|
| HONGGANG CHEN,ET AL: "Single Image Super-Resolution via Adaptive", 《IEEE TRANSACTIONS ON MULTIMEDIA》 * |
| 彭羊平: "基于非负邻域嵌入的单帧图像超分辨率重建算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 李雪玉: "基于非局部相似模型的图像恢复算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109741263A (en) * | 2019-01-11 | 2019-05-10 | 四川大学 | Remote sensed image super-resolution reconstruction algorithm based on adaptive combined constraint |
| CN110020989A (en) * | 2019-05-23 | 2019-07-16 | 西华大学 | A kind of depth image super resolution ratio reconstruction method based on deep learning |
| CN110020989B (en) * | 2019-05-23 | 2022-06-28 | 西华大学 | Depth image super-resolution reconstruction method based on deep learning |
| CN112150354A (en) * | 2019-06-26 | 2020-12-29 | 四川大学 | Single image super-resolution method combining contour enhancement and denoising statistical prior |
| CN112862688A (en) * | 2021-03-08 | 2021-05-28 | 西华大学 | Cross-scale attention network-based image super-resolution reconstruction model and method |
| CN112862688B (en) * | 2021-03-08 | 2021-11-23 | 西华大学 | Image super-resolution reconstruction system and method based on cross-scale attention network |
| CN115115654A (en) * | 2022-06-14 | 2022-09-27 | 北京空间飞行器总体设计部 | Object image segmentation method based on saliency and neighbor shape query |
| CN115115654B (en) * | 2022-06-14 | 2023-09-08 | 北京空间飞行器总体设计部 | Object image segmentation method based on saliency and neighbor shape query |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liang et al. | Details or artifacts: A locally discriminative learning approach to realistic image super-resolution | |
| CN108537734A (en) | Single image super resolution ratio reconstruction method based on gradient profile example dictionary and Weighted adaptive p norms | |
| CN106952228B (en) | Super-resolution reconstruction method of single image based on image non-local self-similarity | |
| Abdelhamed et al. | Ntire 2020 challenge on real image denoising: Dataset, methods and results | |
| Liu et al. | Trident dehazing network | |
| Trinh et al. | Novel example-based method for super-resolution and denoising of medical images | |
| Yu et al. | A unified learning framework for single image super-resolution | |
| CN111709895A (en) | Image blind deblurring method and system based on attention mechanism | |
| Fedorov et al. | Affine non-local means image denoising | |
| Zhang et al. | A single-image super-resolution method based on progressive-iterative approximation | |
| CN112288632B (en) | Single image super-resolution method and system based on simplified ESRGAN | |
| CN111968123B (en) | Semi-supervised video target segmentation method | |
| Cao et al. | New architecture of deep recursive convolution networks for super-resolution | |
| Wang et al. | Gif2video: Color dequantization and temporal interpolation of gif images | |
| CN105513033A (en) | Super-resolution reconstruction method based on non-local simultaneous sparse representation | |
| Makwana et al. | Single image super-resolution via iterative back projection based Canny edge detection and a Gabor filter prior | |
| Li | Image super-resolution using attention based densenet with residual deconvolution | |
| CN114494015A (en) | Image reconstruction method based on blind super-resolution network | |
| CN117575915A (en) | Image super-resolution reconstruction method, terminal equipment and storage medium | |
| Wu et al. | Lightweight asymmetric convolutional distillation network for single image super-resolution | |
| CN113012061A (en) | Noise reduction processing method and device and electronic equipment | |
| Gao et al. | Efficient Multi-scale Network with Learnable Discrete Wavelet Transform for Blind Motion Deblurring | |
| El Fadili et al. | Lapped block image analysis via the method of Legendre moments | |
| Sheta | Restoration of medical images using genetic algorithms | |
| Wei et al. | RSAN: Residual subtraction and attention network for single image super-resolution |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180914 |