CN109544477A - Image denoising algorithm based on self-adapting dictionary study rarefaction representation - Google Patents
Image denoising algorithm based on self-adapting dictionary study rarefaction representation Download PDFInfo
- Publication number
- CN109544477A CN109544477A CN201811410032.3A CN201811410032A CN109544477A CN 109544477 A CN109544477 A CN 109544477A CN 201811410032 A CN201811410032 A CN 201811410032A CN 109544477 A CN109544477 A CN 109544477A
- Authority
- CN
- China
- Prior art keywords
- image
- dictionary
- correlation
- sparse representation
- atoms
- 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
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 claims description 24
- 230000003044 adaptive effect Effects 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims description 8
- 239000000654 additive Substances 0.000 claims description 3
- 230000000996 additive effect Effects 0.000 claims description 3
- 239000000470 constituent Substances 0.000 claims description 2
- 230000000694 effects Effects 0.000 description 9
- 238000012549 training Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 5
- 239000013598 vector Substances 0.000 description 4
- 238000010219 correlation analysis Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 206010028980 Neoplasm Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 201000011510 cancer Diseases 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000556 factor analysis Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013076 uncertainty analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/28—Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/513—Sparse representations
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Processing (AREA)
Abstract
The invention proposes a kind of Image denoising algorithms based on self-adapting dictionary study rarefaction representation, which selects image block according to picture structure similitude first, and constructs the dictionary atom of each image block；Then the grey correlation for calculating atom in different images block, to cluster and select can adapt to the dictionary atom of noise；It finally establishes self-adapting dictionary rarefaction representation and realizes image denoising.By being tested on extensive standard database, inventive algorithm can remove noise well, effectively capture image detail.Grey correlation is applied in dictionary learning rarefaction representation by the present invention, solves the problems, such as automatically selecting for dictionary atom, improves the performance of rarefaction representation；Noise and the removal in image can accurately be told.
Description
Technical Field
The invention belongs to the technical field of computer vision, and particularly relates to an image denoising algorithm based on self-adaptive dictionary learning sparse representation.
Background
The image is usually corrupted by noise during acquisition, recording and transmission, and denoising plays an important role in the field of image processing, and aims to restore the observed noisy image to the original clear image. The three-dimensional block matched filter (BM3D) algorithm [ document 1] (Dabov K, Foi A, Katkovnik V, Egiazarian K (2007). imaging differentiation by space 3-D transform domain imaging transformation on Image Processing,16 (8); 2080-. BM3D reconstructs the images by first finding similar blocks, then superimposes them on the 3D signal, and finally represents the blocks by a typical wavelet thresholding process. However, BM3D contains many parameters such as choice of cardinality, size, switching threshold and similarity measure, which is a complex engineering approach.
In recent years, a machine learning technique based on domain transformation is widely applied to image denoising. The K-singular value decomposition (K-SVD) algorithm proposed by Elad and Aharon [ document 2] (Dabov K, Foi A, Katkovnik V, Egiazarian K (2007). Image differentiating byspase 3-D transform domain formatting. IEEE transformation on Image processing,16(8), 2080-. Dictionary learning is not easy, however, and the denoising model uses vectors instead of original matrices to represent each image block. Zhou et al [ reference 3] (Zhou Mingyuan, Chen Haojun, Paisley John, Ren Lu, Li Lingbo, XingZhongming, Dunson David, Sapiro Guillermo, Carin Lawrence (2012), NonparametricmBaysian dictionary learning for analysis of noise and complex images IEEETransaction on Image Processing,21(1),130-144.) proposed a Beta Process Factor Analysis (BPFA) algorithm, which has a similar denoising effect to K-SVD but a high computational complexity.
Dictionary learning solves the problem of image structure representation, but the traditional dictionary learning method does not consider the intrinsic structure of an image, and how to find an optimal dictionary to reconstruct a high-quality image is still a problem. In recent years, structure sparsity has attracted more and more attention, and researchers try to improve the accuracy of sparse representation by training a structured dictionary to express a hidden structure of data. Dictionary design based on Group Sparse Representation [ document 4] (Suk HI, Wee CY, LeeSW, Shen D (2015). Supervised discrete Group Sparse Representation for Mild Cognitive Impatiention Diagnostics. Neurodynamics, 13(3),277-295.) and structure clustering can solve the Representation problem of image structure. The quality of the learned dictionary, however, depends largely on the training data samples; training samples are closely related, which results in the fact that the learning dictionary is not robust; learning using limited training data such that the dictionary does not contain enough atoms; and the algorithm complexity increases with the number of training samples, large training sets cannot be used in real-time systems. Thus, the choice of dictionary atoms is the key to dictionary learning [ document 5] (Tharmalingam M, Raahemifar K (2013). Sparsity constrained image recovery using nonlinear dictionary with time-shifted OMP signal coding algorithm. proceedings of 26th IEEE cancer Conference on electric and Computer Engineering, 1-5.).
Research on Online Dictionary learning and unsupervised learning is advantageous for providing training data [ document 6] (Lu Cewu, ShiJianping, jiajiaya (2013). Online route Dictionary learning.2013ieee conference on Computer Vision and pattern Recognition, 415-. But depending on the image distribution, it is difficult to obtain uncertain information. On the other hand, the relevance between image blocks is an important criterion, and the strong relevance between the image blocks means that the similar information is contained much. Redundant dictionaries trained by image blocks require more atoms to represent features, and run time increases as the number of atoms increases. Furthermore, the prior information, pixel information, structural information, boundary information, and the like are uncertain, and it is necessary to select effective atoms to represent features and reduce the influence of noise. The conventional correlation analysis method mainly adopts vectors for analysis, and the vector correlation analysis needs to completely know image information, but in most cases, partial image information can be known from a test image, so the vector correlation analysis method has certain limitations. Grey theory [ reference 7] (Deng J L (1989) Introduction to Grey system the journal of Grey System,1(1):1-24.) is one of the uncertainty analysis methods, which requires less training samples and has the advantage of flexible processing of complex scenes.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, selects image blocks by using structural similarity, calculates the correlation among dictionary atoms by using gray correlation and clusters, introduces gray correlation to realize sparse representation, and provides an image denoising algorithm based on self-adaptive dictionary learning sparse representation. The method is realized by the following technical scheme:
the image denoising algorithm based on the adaptive dictionary learning sparse representation comprises the following steps:
step 1) image block division: dividing a test image into different image blocks Ri＝[r，r]Where i is the number of divided image blocks and r is the size of an image block;
step 2) structural similarity calculation: downsampling the test image as a reference image RfCalculating a reference image RfAnd different image areas RiSelecting part of image blocks according to the structural similarity to construct dictionary atoms;
step 3) gray correlation degree calculation, namely calculating the gray correlation degrees ξ of the dictionary atoms in different image areas;
step 4) clustering dictionary atoms, namely rearranging dictionary atoms according to the grey relevance ξ, calculating the grey relevance value from each atom to a clustering center, and clustering to obtain a dictionary D;
step 5) sparse representation: taking the grey correlation as a sparse threshold T, continuously updating the dictionary, and obtaining the sparse representation theta of the imagei；
Step 6), image denoising: obtaining a denoised image using an image reconstruction method defined according to equation (1),
wherein,for de-noising an image, Y is the test image, R is the set of all image regions divided, and λ is the weighting parameter.
The image denoising algorithm based on the adaptive dictionary learning sparse representation is further designed in that the reference image R is calculated according to the formula (2) in the step 2)fAnd different image areas RiStructural similarity between:
wherein,is RfAnd RiThe covariance of (a) of (b),are each Rf、RiStandard deviation of (A), whereinIs RfAnd RiStructural correlation coefficient of (a);are each Rf、RiThe average value of (a) of (b),has a value range of [0, 1]]Measured at RfAnd RiDifference in brightness between, if and only ifWhen it is 1;to measure the contrast between a reference image and an image area.
The image denoising algorithm based on the adaptive dictionary learning sparse representation is further designed in that, in the step 2), part of image blocks are selected according to the structural similarity to construct dictionary atoms: in the calculation of the reference image RfAnd different image areas RiAnd setting the intermediate value of the structural similarity as the threshold of the selected image block while constructing the dictionary atom of each image block.
The image denoising algorithm based on the self-adaptive dictionary learning sparse representation is further designed in that the adjacent image blocks RiAnd Ri+1The constituent dictionary atoms can be represented as two sequences p, respectivelyiAnd pi+1：
pi:＝{pi(1),pi(2),pi(3),...,pi(r2)} (3)
pi+1:＝{pi+1(1),pi+1(2),pi+1(3),...,pi+1(r2)} (4)
The gray correlation between two image block dictionary atoms is calculated as follows:
wherein η is a constant number at (0, 1)]K is 1 to r2Is constant.
The image denoising algorithm based on the self-adaptive dictionary learning sparse representation is further designed in that the dictionary atoms are rearranged according to the gray correlation values obtained through calculation in the step 4), the centers of the dictionary atoms are initialized randomly, the gray correlation from each atom to the clustering center is calculated, and the clustering center with the maximum gray correlation degree with the dictionary atoms is divided into the most relevant classes; then calculating the grey correlation degree among all atoms in the most relevant class to obtain a new clustering center; and finally, setting a stopping rule according to the difference of the correlation values between the classes to obtain a dictionary D.
The image denoising algorithm based on the adaptive dictionary learning sparse representation is further designed in that the threshold value T relates to the variance of white Gaussian noise, an image containing additive white Gaussian noise is denoised according to the formula (6),
Y＝X+n (6)
wherein Y is a test image, X is an original image, and n is white Gaussian noise;
in order to make the algorithm adaptive, the image noise level is represented by the degree of correlation between image blocks, and the correlation of the whole image is calculated according to equation (7):
where l is the size of the sparse dictionary, ξhRepresenting the gray associated features of the entire image.
The image denoising algorithm based on the adaptive dictionary learning sparse representation is further designed in that the sparse threshold T and the gray correlation ξhIs defined as formula (8):
T×ξh＝C (8)
wherein C is a constant;
threshold T is associated with value ξ through GrayhInstead, the dictionary is continuously updated, and a sparse representation of the image is obtained according to equation (9):
wherein, YiIth test sample, ΘiIs a sparse representation of the ith sample.
The invention has the beneficial effects that:
according to the method, redundant atoms in the over-complete dictionary can be reduced through dictionary learning, and image structure information can be reserved through structural similarity; the grey correlation values are used as a rule for selecting dictionary atoms to adapt to noise levels and image structure characteristics. The method can accurately distinguish information and noise in the image, thereby reducing the pseudo-Gibbs effect, storing important information in the image and having superior performance in denoising.
Drawings
FIG. 1 is a flow diagram of dictionary atom selection in accordance with the present invention.
FIG. 2 is a flow chart of dictionary atom clustering in accordance with the present invention.
FIG. 3 is a graph of gray correlation values for different noise variations according to the present invention.
FIG. 4 noise figure used in the experiments of the present invention.
FIG. 5 is a denoising effect graph of the present invention using the K-SVD method.
FIG. 6 is a denoising effect graph of the present invention using the BM3D method.
FIG. 7 is a diagram of the denoising effect of the method of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
The image denoising algorithm based on the adaptive dictionary learning sparse representation of the embodiment utilizes the adaptive dictionary sparse representation to realize image denoising on the basis of using gray correlation to select dictionary atoms aiming at a noise image; the specific implementation comprises the following steps:
step 1): image block division: dividing a test image into different image blocks Ri＝[r，r]Where i is the number of divided image blocks, r is the size of the image block, r is 8;
step 2): the method comprises the following steps of taking a detected image after down-sampling as a reference image, reducing the size and noise density of the image, and calculating the structural similarity between the reference image and different image blocks:
whereinIs a reference image RfAnd image block RiThe covariance of (a) of (b),andis the standard deviation, where the first multiplier is RfAnd RiStructural correlation coefficient of (a);is the average value of the image blocks,has a value range of [0, 1]]Measured at RfAnd RiDifference in brightness between, if and only ifWhen it is 1; the third multiplier measures the contrast between the reference image and the image block. Meanwhile, setting the intermediate value of the structural similarity as a threshold value for selecting the image block, and constructing a dictionary atom of each image block, as shown in fig. 1.
Step 3) calculating the grey correlation degree ξ of dictionary atoms in different image blocks, and setting RiAnd Ri+1For adjacent image blocks, each image block dictionary atom may form two sequences pi，pi+1：
pi:＝{pi(1),pi(2),pi(3),...,pi(r2)} (9)
pi+1:＝{pi+1(1),pi+1(2),pi+1(3),...,pi+1(r2)} (10)
The gray correlation ξ between the two sequences is calculated as follows:
wherein η is a constant number at (0, 1)]K is 1 to r2Is constant.
Step 4): selecting the gray correlation value obtained by calculation in the step 3) as a clustering principle, and selecting effective dictionary atoms to be classified into the most relevant classes; selecting a gray level correlation value, and setting atoms as the most relevant classes according to the relationship among the atoms; then, calculating the grey correlation degrees among all atoms again to obtain a new clustering center; and setting a stopping rule according to the difference of the relation values between the classes to finally obtain a dictionary D, which is shown in figure 2.
Step 5): the noise is unknown in the denoising process, and the image containing the additive white Gaussian noise is denoised.
Y＝X+n (12)
Where Y is the test image, X is the original image, and n is white Gaussian noise. In order to adapt the algorithm, the noise of the image is represented by the correlation between the image blocks. To solve the noise estimation problem, the global estimation and the local estimation are considered, and the relation between the whole images is calculated:
where l is the size of the sparse dictionary, ξhGray associated features representing the entire imageIn general, as the noise variance increases, the gray correlation values decrease approximately, see FIG. 3, so the sparse threshold T and gray correlation ξhThe relationship of (c) is defined as:
T×ξh＝C (14)
where C is a constant, the variance of the noise in the test image is represented by a gray correlation value, i.e., the sparse threshold T may be represented by a gray correlation value ξhReplacing, continuously updating the dictionary, and obtaining sparse representation of the image:
wherein Y isiIth test sample, ΘiIs a sparse representation of the ith sample.
Step 6): obtaining a denoised picture by using an image reconstruction method, wherein the image reconstruction method can be defined as the following steps:
whereinFor de-noising an image, Y is a test image, R is a set of all divided image blocks, and λ is a weight parameter.
The inventor of the present application has experimentally verified the effect of the method of the present invention, and fig. 4 is a noisy image, which is used as the test image of the present embodiment, and the denoising effect using the K-SVD method and the BM3D method is as shown in fig. 5 and fig. 6; the algorithm of the invention combines the image structure characteristics to obtain the image dictionary and sparse representation of gray cluster, and realizes denoising with the effect as shown in FIG. 7.
The inventor of the application also compares different types of image denoising methods, and adopts a peak signal-to-noise ratio (PSNR) as a quantitative comparison standard for different noise variables sigma contained in the image. A larger PSNR value means that the image contains less noise, and the experimental results are shown in the following table.
TABLE 1 image denoising Effect of different algorithms
As can be seen from the table, the algorithm of the present invention outperforms other algorithms at different noise levels. The experimental apparatus is configured to: windows 7 operating system, 3.3Ghz master frequency inter (R) core (TM) i3 CPU. According to the method, grey correlation is applied to dictionary learning sparse representation, dictionary atoms can be automatically selected according to different noise levels, and the performance of sparse representation is improved; the noise in the image can be accurately distinguished and removed.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (7)
1. An image denoising algorithm based on adaptive dictionary learning sparse representation is characterized by comprising the following steps:
step 1) image block division: dividing a test image into different image blocks Ri＝[r，r]Where i is the number of divided image blocks and r is the size of an image block;
step 2) structural similarity calculation: downsampling the test image as a reference image RfCalculating a reference image RfAnd different image areas RiStructural similarity between them, according to structural similaritySelectively selecting part of image blocks to construct dictionary atoms;
step 3) gray correlation degree calculation, namely calculating the gray correlation degrees ξ of the dictionary atoms in different image areas;
step 4) clustering dictionary atoms, namely rearranging dictionary atoms according to the grey relevance ξ, calculating the grey relevance value from each atom to a clustering center, and clustering to obtain a dictionary D;
step 5) sparse representation: taking the grey correlation as a sparse threshold T, continuously updating the dictionary, and obtaining the sparse representation theta of the imagei；
Step 6), image denoising: obtaining a denoised image using an image reconstruction method defined according to equation (1),
wherein,for de-noising an image, Y is the test image, R is the set of all image regions divided, and λ is the weighting parameter.
2. The image denoising algorithm based on adaptive dictionary learning sparse representation according to claim 1, wherein the reference image R is calculated according to formula (2) in the step 2)fAnd different image areas RiStructural similarity between:
wherein,is RfAnd RiThe covariance of (a) of (b),are each Rf、RiStandard deviation of (A), whereinIs RfAnd RiStructural correlation coefficient of (a); are each Rf、RiThe average value of (a) of (b),has a value range of [0, 1]]Measured at RfAnd RiDifference in brightness between, if and only ifWhen it is 1;to measure the contrast between a reference image and an image area.
3. The image denoising algorithm based on adaptive dictionary learning sparse representation according to claim 2, wherein the selecting part of the image blocks to construct dictionary atoms according to the structural similarity in the step 2) is: in the calculation of the reference image RfAnd different image areas RiAnd setting the intermediate value of the structural similarity as the threshold of the selected image block while constructing the dictionary atom of each image block.
4. The adaptive dictionary learning sparse representation-based image denoising algorithm of claim 3, wherein the adjacent image blocks RiAnd Ri+1The constituent dictionary atoms can be represented as two sequences p, respectivelyiAnd pi+1：
pi:＝{pi(1),pi(2),pi(3),...,pi(r2)} (3)
pi+1:＝{pi+1(1),pi+1(2),pi+1(3),...,pi+1(r2)} (4)
The gray correlation between two image block dictionary atoms is calculated as follows:
wherein η is a constant number at (0, 1)]K is 1 to r2Is constant.
5. The image denoising algorithm based on adaptive dictionary learning sparse representation according to claim 1, wherein the dictionary atoms are rearranged according to the calculated gray correlation value in the step 4), the centers of the dictionary atoms are initialized randomly, the gray correlation from each atom to the cluster center is calculated, and the cluster center with the largest gray correlation degree with the dictionary atoms is classified into the most relevant class; then calculating the grey correlation degree among all atoms in the most relevant class to obtain a new clustering center; and finally, setting a stopping rule according to the difference of the correlation values between the classes to obtain a dictionary D.
6. The adaptive dictionary learning sparse representation-based image denoising algorithm of claim 1, wherein the threshold T relates to variance of white Gaussian noise, an image containing additive white Gaussian noise is denoised according to equation (6),
Y＝X+n (6)
wherein Y is a test image, X is an original image, and n is white Gaussian noise;
in order to make the algorithm adaptive, the image noise level is represented by the degree of correlation between image blocks, and the correlation of the whole image is calculated according to equation (7):
where l is the size of the sparse dictionary, ξhRepresenting the gray associated features of the entire image.
7. The adaptive dictionary learning sparse representation-based image denoising algorithm of claim 6, wherein the sparse threshold T is associated ξ with grayhIs defined as formula (8):
T×ξh＝C (8)
wherein C is a constant;
threshold T is associated with value ξ through GrayhInstead, the dictionary is continuously updated, and a sparse representation of the image is obtained according to equation (9):
wherein, YiIth test sample, ΘiIs a sparse representation of the ith sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811410032.3A CN109544477A (en) | 2018-11-23 | 2018-11-23 | Image denoising algorithm based on self-adapting dictionary study rarefaction representation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811410032.3A CN109544477A (en) | 2018-11-23 | 2018-11-23 | Image denoising algorithm based on self-adapting dictionary study rarefaction representation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109544477A true CN109544477A (en) | 2019-03-29 |
Family
ID=65849484
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811410032.3A Pending CN109544477A (en) | 2018-11-23 | 2018-11-23 | Image denoising algorithm based on self-adapting dictionary study rarefaction representation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109544477A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110213462A (en) * | 2019-06-13 | 2019-09-06 | Oppo广东移动通信有限公司 | Image processing method, device, electronic equipment and image processing circuit |
| CN110176029B (en) * | 2019-04-29 | 2021-03-26 | 华中科技大学 | Image restoration and matching integrated method and system based on level sparse representation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184526A (en) * | 2011-04-22 | 2011-09-14 | 西安电子科技大学 | Natural image denoising method based on dictionary learning and block matching |
| CN103136719A (en) * | 2013-03-22 | 2013-06-05 | 南通大学 | Image dictionary construction method |
| CN105279740A (en) * | 2015-11-17 | 2016-01-27 | 华东交通大学 | Image denoising method based on sparse regularization |
| CN106204483A (en) * | 2016-07-08 | 2016-12-07 | 桂林电子科技大学 | Multiplicative noise minimizing technology based on non local self-adapting dictionary |
| CN108537752A (en) * | 2018-03-30 | 2018-09-14 | 广东工业大学 | Image processing method based on non local self-similarity and rarefaction representation and device |
-
2018
- 2018-11-23 CN CN201811410032.3A patent/CN109544477A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184526A (en) * | 2011-04-22 | 2011-09-14 | 西安电子科技大学 | Natural image denoising method based on dictionary learning and block matching |
| CN103136719A (en) * | 2013-03-22 | 2013-06-05 | 南通大学 | Image dictionary construction method |
| CN105279740A (en) * | 2015-11-17 | 2016-01-27 | 华东交通大学 | Image denoising method based on sparse regularization |
| CN106204483A (en) * | 2016-07-08 | 2016-12-07 | 桂林电子科技大学 | Multiplicative noise minimizing technology based on non local self-adapting dictionary |
| CN108537752A (en) * | 2018-03-30 | 2018-09-14 | 广东工业大学 | Image processing method based on non local self-similarity and rarefaction representation and device |
Non-Patent Citations (4)
| Title |
|---|
| "基于稀疏表示和自适应字典学习的"高分一号"遥感图像去噪", 光电工程 * |
| HONGJUN LI等: "Automatic Dictionary Learning Sparse Representation for Image Denoising", JOURNAL OF GREY SYSTEM * |
| ZHOU WANG等: "A Universal Image Quality Index", IEEE SIGNAL PROCESSING LETTERS * |
| 李洪均;谢正光;胡伟;王伟;: "字典原子优化的图像稀疏表示及其应用", 东南大学学报(自然科学版) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110176029B (en) * | 2019-04-29 | 2021-03-26 | 华中科技大学 | Image restoration and matching integrated method and system based on level sparse representation |
| CN110213462A (en) * | 2019-06-13 | 2019-09-06 | Oppo广东移动通信有限公司 | Image processing method, device, electronic equipment and image processing circuit |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109360156B (en) | Single image rain removing method based on image block generation countermeasure network | |
| CN108921800B (en) | Non-local mean denoising method based on shape self-adaptive search window | |
| CN110175649B (en) | Rapid multi-scale estimation target tracking method for re-detection | |
| CN113191979B (en) | Non-local mean denoising method for partitioned SAR (synthetic aperture radar) image | |
| CN107657593A (en) | A kind of removing rain based on single image method | |
| CN107944497A (en) | Image block method for measuring similarity based on principal component analysis | |
| CN104657951A (en) | Multiplicative noise removal method for image | |
| Onur | Improved image denoising using wavelet edge detection based on Otsu’s thresholding | |
| Qi et al. | Research on the image segmentation of icing line based on NSCT and 2-D OSTU | |
| CN112991199A (en) | Image high-low frequency decomposition noise removing method based on residual error dense network | |
| CN117058735A (en) | Micro-expression recognition method based on parameter migration and optical flow feature extraction | |
| CN112598708A (en) | Hyperspectral target tracking method based on four-feature fusion and weight coefficient | |
| CN109544477A (en) | Image denoising algorithm based on self-adapting dictionary study rarefaction representation | |
| CN116823642A (en) | Image motion blur removing method and device, electronic equipment and medium | |
| CN115272113A (en) | Image deblurring method based on multi-scale frequency separation network | |
| CN116385312A (en) | Low-illumination image denoising method based on phase correlation | |
| CN111047537A (en) | System for recovering details in image denoising | |
| Xie et al. | DHD-Net: A novel deep-learning-based dehazing network | |
| CN112927169B (en) | Remote sensing image denoising method based on wavelet transformation and improved weighted kernel norm minimization | |
| Rani et al. | ELM-Based Shape Adaptive DCT Compression technique for underwater image compression | |
| Liu et al. | The validity analysis of the non-local mean filter and a derived novel denoising method | |
| Yapici et al. | A review of image denoising with deep learning | |
| CN110335225B (en) | Method for fusing infrared light image and visible light image | |
| CN108416756B (en) | Regional perception image denoising method based on machine learning | |
| Zhang et al. | A Single-Stage Unsupervised Denoising Low-Illumination Enhancement Network Based on Swin-Transformer |
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: 20190329 |