CN106504207B - A kind of image processing method - Google Patents
A kind of image processing method Download PDFInfo
- Publication number
- CN106504207B CN106504207B CN201610948033.8A CN201610948033A CN106504207B CN 106504207 B CN106504207 B CN 106504207B CN 201610948033 A CN201610948033 A CN 201610948033A CN 106504207 B CN106504207 B CN 106504207B
- Authority
- CN
- China
- Prior art keywords
- matrix
- image
- denoising
- image matrix
- noise
- 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.)
- Active
Links
- 238000003672 processing method Methods 0.000 title abstract description 15
- 239000011159 matrix material Substances 0.000 claims abstract description 180
- 238000000034 method Methods 0.000 claims abstract description 44
- 239000013598 vector Substances 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims description 23
- 238000001914 filtration Methods 0.000 claims description 14
- 238000000354 decomposition reaction Methods 0.000 claims description 7
- 238000012804 iterative process Methods 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 abstract description 10
- 230000003044 adaptive effect Effects 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 23
- 238000012360 testing method Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 8
- 241000196324 Embryophyta Species 0.000 description 7
- 238000011156 evaluation Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000011160 research Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000007796 conventional method Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 240000004760 Pimpinella anisum Species 0.000 description 1
- 241000758706 Piperaceae Species 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000000087 stabilizing effect Effects 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000011426 transformation method 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
The embodiment of the present invention provides a kind of image processing method, which comprises carries out total variation to noise image matrix, obtains the first denoising image array；According to the first denoising image array and the noise image matrix, residual image matrix is obtained；The residual image matrix is subjected to adaptive wiener filter, obtains filtered residual image matrix；The first denoising image array, the filtered residual image matrix and weight vectors are subjected to second of denoising according to the first preset rules, obtain the second denoising image array.The method takes full advantage of the prior information of image, to preferably remain edge, the details of image, while obtaining high s/n ratio, structural similarity also keeps higher level, sufficiently meets people's visual effect while favorably removal noise.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to an image data processing method.
Background
The image signal is inevitably interfered by noise in the process of acquiring, transmitting and storing the image signal, so that the information of the image is submerged, and the visual quality of the image is seriously influenced. A great deal of image edges and detailed features are submerged, which brings great difficulty to the analysis and subsequent processing of the image. The elimination of image noise is an important research content in image preprocessing. And a good foundation is provided for subsequent processing such as edge detection, image segmentation, feature extraction and pattern recognition. Therefore, how to effectively remove noise while maintaining the definition of image details and the contrast of the image becomes a hot research focus of people.
With the introduction of non-local ideas, people begin to change from self-similarity of images to attention on image structure. Among them, Alessandro Foi, Giacomo Boracchi proposes a funnel-based non-local self-similarity algorithm (Anis. Fov. NL-Means). The basic idea is as follows: based on HVS space, concave characters are adopted to replace calculation of similar distances in traditional non-local denoising, and radial anisotropic concave operators are used to increase fidelity of structures and edges. The combination of the space domain and the transform domain, when the structure of the image is converted into the structure of the pixel matrix, the rank of the matrix is used as an index for measuring the correlation of matrix columns or rows, and the structural information of the similar matrix can be well described. Video clips captured by a still camera have a well-defined low rank property on which background modeling and foreground extraction can be performed. This also indicates that the matrix formed by the locally similar patches under the natural image is low rank and can be used for the recovery of high performance images. Therefore, it is a research focus to recover potential image information from the degradation model of the image matrix, i.e. the approximation of the low rank matrix. Due to the rapid development of under-convex and non-convex techniques, some low rank matrix approximations have been studied in recent years, as well as many important models and algorithms.
Low rank matrix approximations can generally be divided into two categories: low rank matrix decomposition (LRMF) and Nuclear Norm Minimization (NNM). LRMF aims to find a matrix X that is as close to Y as possible on some data fidelity function, given the matrix Y. And can also be decomposed into the product of two low rank matrices. Many algorithms based on LRMF have been proposed, including a stabilizing algorithm from classical singular value decomposition to many L1-norm. NNM is another form of low rank matrix approximation that differs from LRMF in that the approximation matrix X is found while it is minimized to the kernel standard. Also, NNM is advantageous in that it is convex in fidelity term specific data problems, while LRMF is non-convex, thus attracting tremendous research interest to scholars in recent years. Candes and Recht demonstrated that most low rank matrices can be recovered by an NNM problem; caietal demonstrates that the problem of singular values of vibrational operation with fidelity to F-norm data can be easily solved with NNM. Although NNM has been widely used for low rank matrix approximation, it still has some problems, and in order to ensure convexity, the prior information of the image is ignored, and the edge details of the image cannot be preserved.
Disclosure of Invention
It is therefore an object of the present invention to provide an image data processing method to solve the above problems.
In order to achieve the above purpose, the embodiment of the present invention adopts the following technical solutions:
the embodiment of the invention provides an image data processing method, which comprises the following steps: carrying out total variation on the noise image matrix to obtain a first denoising image matrix; acquiring a residual image matrix according to the first denoising image matrix and the noise image matrix; performing self-adaptive wiener filtering on the residual image matrix to obtain a filtered residual image matrix; and carrying out secondary denoising treatment on the first denoised image matrix, the filtered residual image matrix and the weight vector according to a first preset rule to obtain a second denoised image matrix.
Compared with the prior art, the image data processing method provided by the embodiment of the invention has the advantages that the noise image matrix is subjected to total variation firstly, the residual image matrix is subjected to self-adaptive wiener filtering, the first denoising image matrix, the filtered residual image matrix and the weight vector are subjected to secondary denoising treatment according to the first preset rule, the second denoising image matrix is obtained, and the prior information of the image is fully utilized, so that the noise is favorably removed, the edge and the detail of the image are better kept, the structural similarity is kept at a higher level while a high signal-to-noise ratio is obtained, and the visual effect of people is fully met.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
Fig. 1 is a block diagram of a server according to an embodiment of the present invention.
Fig. 2 is a flowchart of an image data processing method according to a first embodiment of the present invention.
Fig. 3 is a flowchart of an image data processing method according to a second embodiment of the present invention.
Fig. 4 is a block diagram of an image data processing apparatus according to a third embodiment of the present invention.
FIG. 5 is a comparison graph of the noise removal effect of the third embodiment of the present invention on the lena256X256 test chart compared with the conventional method (AFNM, BM3D, WNNM).
FIG. 6 is a comparison graph of the denoising effect of the Monarch256X256 test chart according to the third embodiment of the present invention and the prior art (AFNM, BM3D, WNNM).
FIG. 7 is a comparison graph of the denoising effect of the third embodiment of the present invention on peppers256X256 test chart compared with the prior art (AFNM, BM3D, WNNM).
FIG. 8 is a comparison graph of the noise removal effect of the third embodiment of the present invention and the existing method (BM3D, WNNM) on a plant monitoring test chart.
Fig. 9 is a comparison graph of the effect of the third embodiment of the present invention on monitoring the noise level of a plant compared to the prior art method (BM3D, WNNM).
FIG. 10 is a comparison of the third embodiment of the present invention with the prior art method (BM3D, WNNM) for IQA evaluation in a plant monitoring system.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures. Meanwhile, in the description of the present invention, the terms "first", "second", and the like are used only for distinguishing the description, and are not to be construed as indicating or implying relative importance.
As shown in fig. 1, there is a block schematic diagram of a server 200. The server 200 includes a memory 201, a processor 202, and a network module 203.
The memory 201 may be used to store software programs and modules, such as program instructions/modules corresponding to the image data processing method and apparatus in the embodiments of the present invention, and the processor 202 executes various functional applications and data processing by running the software programs and modules stored in the memory 201, that is, implementing the application topic recommendation method in the embodiments of the present invention. Memory 201 may include high speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. Further, the software programs and modules in the memory 201 may further include: an operating system 221 and a service module 222. The operating system 221, which may be LINUX, UNIX, WINDOWS, for example, may include various software components and/or drivers for managing system tasks (e.g., memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide an operating environment for other software components. The service module 222 runs on the basis of the operating system 221, and monitors a request from the network through the network service of the operating system 221, completes corresponding data processing according to the request, and returns a processing result to the client. That is, the service module 222 is used to provide network services to clients.
The network module 203 is used for receiving and transmitting network signals. The network signal may include a wireless signal or a wired signal.
It is to be understood that the configuration shown in fig. 1 is merely illustrative and that the server 200 may include more or fewer components than shown in fig. 1 or may have a different configuration than shown in fig. 1. The components shown in fig. 1 may be implemented in hardware, software, or a combination thereof. In addition, the server in the embodiment of the present invention may further include a plurality of servers with different specific functions.
Fig. 2 is a flowchart illustrating an image data processing method according to a first embodiment of the present invention, and referring to fig. 2, this embodiment describes a processing flow of a server, where the method includes:
step S310, carrying out total variation on the noise image matrix to obtain a first denoising image matrix.
The total variation means that the image denoising is modeled into a minimization problem of an energy function, so that the image reaches a smooth state, the anisotropic diffusion equation of the partial differential equation is used for processing the noise image, the edge can be maintained while the noise is smoothed, and the contradiction between the image detail recovery and the noise suppression is better solved.
Step S320, obtaining a residual image matrix according to the first denoising image matrix and the noise image matrix.
As an embodiment, a matrix of the noise image matrix minus the first denoised image matrix may be used as the residual image matrix.
And step S330, performing self-adaptive wiener filtering on the residual image matrix to obtain a filtered residual image matrix.
Compared with other filters, the self-adaptive wiener filtering has better filtering effect and good selectivity, and can better retain the edge and high-frequency details of an image. Therefore, the wiener filtering can well inhibit noise while keeping the structural information of image loss on the residual image.
Step S340, performing a second denoising process on the first denoised image matrix, the filtered residual image matrix, and the weight vector according to a first preset rule, to obtain a second denoised image matrix.
As an embodiment, the first preset rule is a weighted nuclear norm model constructed according to a superposition matrix formed by adding the first denoised image matrix and the filtered residual image matrix, a low-rank matrix corresponding to the superposition matrix, a weight vector, a first parameter and a second parameter.
The low-rank matrix corresponding to the superposition matrix is an image information low-rank matrix corresponding to the superposition matrix, and the image information is a pollution-free image composed of pure image textures, structures and the like.
Further, the weighted kernel norm model is:
wherein, YiIs a superposition matrix formed by adding the first denoised image matrix and the filtered residual image matrix, Xi is a low rank matrix corresponding to the superposition matrix, λ is a first parameter, σ n is a second parameter,w＝[w1,w2.....wi]in order to be a weight vector, the weight vector,σi(Yi) Is YiI-th singular value of (a), i-1, 2.
As one of the ways, the second parameter σ n is a noise variance value. The first parameter λ is a soft threshold adjustment factor.
The weight vector w ═ w1,w2.....wi]The calculation formula of each element is as follows:
where c is a constant greater than zero, n is a third parameter, and ε is 10-16Is a fixed value and is used as a reference,σi(Yi) Is YiN, Y, i ═ 1,2iIs a superposition matrix formed by adding the first denoised image matrix and the filtered residual image matrix.
According to the image data processing method provided by the embodiment of the invention, the noise image matrix is subjected to total variation firstly, the residual image matrix is subjected to self-adaptive wiener filtering, and the first de-noised image matrix, the filtered residual image matrix and the weight vector are subjected to secondary de-noising treatment according to a first preset rule to obtain a second de-noised image matrix, wherein the second de-noised image matrix fully utilizes the prior information of the image, so that the noise is removed favorably, the edges and details of the image are better kept, the structural similarity is kept at a higher level while a high signal-to-noise ratio is obtained, and the visual effect of people is fully met.
Fig. 3 is a flowchart illustrating an image data processing method according to a second embodiment of the present invention, and referring to fig. 3, this embodiment describes a processing flow of a server, where the method includes:
step S410, carrying out total variation on the noise image matrix to obtain a first denoising image matrix.
Step S420, obtaining a residual image matrix according to the first denoising image matrix and the noise image matrix.
And step S430, performing self-adaptive wiener filtering on the residual image matrix to obtain a filtered residual image matrix.
Step S440, adding the first denoised image matrix and the filtered residual image matrix to form a superposition matrix.
Step S450, dividing the superposition matrix into a plurality of sub-matrices according to the noise variance value.
The superposition matrix may be divided according to different sizes according to the difference of the noise variance values, for example, when the noise variance values are respectively σn≤25，25<σn≤40，40<σn≤60，60<σnIn this case, the size correspondence of the sub-matrices may be set to 90 × 90, 90 × 90, 130 × 130, 140 × 140, respectively.
Step S460, according to a second preset rule, obtaining a similar matrix corresponding to each sub-matrix respectively.
As an implementation manner, the sub-matrices may be further divided according to different noise variance values, for example, when the noise variance values are respectively σ, the sub-matrices may be divided into different sizesn≤25，25<σnIn this case, the sub-matrix correspondence may be divided into 35 × 35 and 40 × 40, respectively, to obtain a small-sized matrix. And after further dividing the sub-matrixes, searching and matching according to the divided small-specification matrixes to obtain similar matrixes corresponding to the sub-matrixes.
And the second preset rule is regularization iterative processing.
As an embodiment, the regularization iterative formula of the regularization iterative process isWherein,y(0)y is a superposition matrix formed by superposing the first denoised image matrix and the filtered residual image matrix, and delta is an iteration step parameter.
And the iteration times set by k are 6, 8,9 and 11 in sequence according to different noise equation values.
When the iteration step parameter δ is 0.1, the soft threshold adjustment factor λ is 0.56 when the noise variance value is smaller than equal to 40, and the soft threshold adjustment factor λ is 0.58 when the noise variance value is larger than 40.
Step S470, performing singular value decomposition on each of the similar matrices, respectively, to obtain a first matrix, a singular value diagonal matrix, and a second matrix corresponding to each of the similar matrices.
Assuming that the similarity matrix yi is an m × n order matrix, there is a decomposition such that
[U,Σ,V]＝SVD(yi)＝UΣV
Wherein the first matrix U is a unitary matrix of order mxm; Σ is an mxn-order singular value diagonal matrix; the second matrix V is a unitary matrix of order n × n. Such a decomposition is called the singular value decomposition of yi. The element Σ i, i on the Σ diagonal is the singular value of yi.
Step S480, multiplying the singular value diagonal matrix corresponding to each similar matrix and the weight vector, respectively, to obtain a third matrix corresponding to each similar matrix.
As an embodiment, the weight vector w ═ w1,w2.....wi]Each element inWhere c is a constant greater than zero, n is a third parameter, and ε is 10-16Is a fixed value and is used as a reference,σi(Yi) Is YiN, Yi is a superposition matrix formed by superposing the first denoised image matrix and the filtered residual image matrix.
Further, each element in the third matrix may be calculated according to the following formula:
Sw(Σ)ii＝max(Σii-wi,0)，i＝1,2......n
wherein, said ∑ isiiIs the value of an element in the diagonal matrix of singular values, wiIs the value of an element in the weight vector.
Step S490, multiplying the first matrix, the third matrix, and the transposed matrix of the second matrix corresponding to each similar matrix, respectively, to obtain a second denoising submatrix corresponding to each similar matrix.
The calculation can be made according to the following formula:
where U is the first matrix, Sw(Σ) is the third matrix and V is the second matrix.
And S500, summing all the second denoising submatrices to obtain a second denoising image matrix.
Further, in order to illustrate the beneficial effects of the embodiment of the present invention, standard Lenas256X256, Manchar256X256, and peppers256X256 images commonly used in image denoising are adopted, and zero-mean additive white gaussian noise with standard deviation σ of 10, 25,50,70, and 100 is added to the three images, so as to generate 21 noisy graphs as test data in total. Respectively using (1) three-dimensional block matching denoising, wherein the method is abbreviated as BM 3D; (2) self-similar non-local denoising of a funnel, wherein the method is called AFNM for short; (3) denoising the weighted nuclear norm, wherein the method is abbreviated as WNNM; (4) the image data processing method in the embodiment of the invention, wherein the threshold value of the self-adaptive wiener filtering is set to be 3X3, the parameter c in the second denoising treatment adopts 2.8284, and the parameter c is processed according to different noise variance values sigman≤25、25<σn≤40、40<σn≤60、60<σnThe search windows are set to 7 × 7, 8 × 8, and 9 × 9 in this order, and the number of iterations K is set to 6, 8,9, and 11 in this order. SigmanThe size of the non-local search window is set to 35 when the value is less than or equal to 25, and the rest are 40. Meanwhile, a frame of a monitoring video image of a certain factory building is processed and compared by the methods (1), (2) and the text.
The evaluation of the image denoising effect is divided into two categories, namely subjective evaluation criteria and objective evaluation criteria. The subjective standard is mainly that images are directly observed through the vision of human eyes, so that the image quality is evaluated, the image quality is good, the denoising effect is good if the image is clear, and otherwise, the denoising effect is poor. Subjectively, the embodiment of the invention adopts the Structural Similarity (SSIM) to measure the denoising effect of the image, and objectively, the embodiment of the invention adopts the peak signal-to-noise ratio (PSNR) to measure the denoising effect of the image. In real life, no noise-free image exists, and the evaluation standards of the denoised image, such as PSNR (signal to noise ratio) and SSIM (structural similarity) will be invalid. The invention adopts an evaluation method combining the ambiguity and the noise level, measures the ambiguity by using the average edge width, and represents the noise degree by using the noise point information of a smooth region, which is called IQA for short, and the smaller the IQA, the better the image quality.
The peak snr results for Lena, Manchar, and peppers images with noise standard deviations σ of 10, 25,50,70, and 100, respectively, by the four denoising methods are shown in table 1. Table 1 different denoising methods are used to perform peak signal-to-noise ratios PSNR and SSIM on different images under different noise intensities.
TABLE 1
From table 1, it can be seen that the method of this embodiment is higher than BM3D and AFLM in both signal-to-noise ratio and structural similarity, and especially shows superiority under high noise. Compared with the original weighted kernel norm denoising algorithm, although the processing result of lena is slightly worse in signal-to-noise ratio, it is better in SSIM. Satisfactory results are obtained for the processing of other two images, which shows that the invention can effectively improve the denoising effect of the images and meet the feeling of the human visual system.
FIG. 5 is a graph showing the comparison of the third embodiment of the present invention with the denoising effect of the existing method (AFNM, BM3D, WNNM) on the lena256X256 test chart. FIG. 6 is a graph showing the comparison of the denoising effect of the third embodiment of the present invention on the Monarch256X256 test chart with the existing method (AFNM, BM3D, WNNM). FIG. 7 is a comparison graph of the denoising effect of the third embodiment of the present invention on peppers256X256 test chart with the conventional method (AFNM, BM3D, WNNM).
Fig. 5, 6 and 7 show the comparison of the denoising effects of the above four methods when the noise standard deviation is 25,50,70 and 100. As can be seen from fig. 5, 6, and 7, AFNM denoising and BM3D denoising can eliminate noise to a certain extent, but BM3D denoising well retains the detail edges of an image at a high noise level, but introduces a large number of false stripes in the detail texture, and the AFNM denoised image has a serious mottling effect, so that the processed image has a large distortion. By adopting WNNM denoising, the visual distortion caused by the phenomenon of false stripes and mottling can be effectively solved, noise can be well removed, the integrity of edges and textures is kept, and more or less false information appears. The WNNM-based secondary image denoising method is adopted, the advantages of the WNNM are fully adopted, iteration times are reduced, good visual experience is brought, and although the signal to noise ratio of lena processing is slightly worse than that of the WNNM, the SSIM is improved as a whole. It can be seen that the visual quality of the image obtained by the WNNM-based secondary image denoising method is better than that of the image obtained by other transformation methods.
FIG. 8 is a comparison graph of the noise removal effect of the third embodiment of the present invention and the existing method (BM3D, WNNM) on a plant monitoring test chart. Fig. 9 is a comparison graph of the third embodiment of the present invention and the effect of the prior art method (BM3D, WNNM) on monitoring the noise level of a certain plant, where the abscissa in fig. 9 is the noise variance value and the ordinate is the noise level. Fig. 10 is a comparison graph of the third embodiment of the present invention and the conventional method (BM3D, WNNM) for the IQA evaluation of a plant monitoring test, where the abscissa in fig. 10 is the noise variance value and the ordinate is IQA.
Comparing fig. 8,9 and 10, the denoising processing effects of BM3D, AFNM and the embodiment of the present invention on the plant monitoring image are shown. From fig. 8, it is seen that the three algorithms are all effective in removing noise, but the visual effect of BM3D is clearly not as good as the other two. From fig. 9 and fig. 10, it can be seen that the denoising effect is best when σ is 36, the noise level is only 0.0034, IQA is 1.3607, which is better than the other two methods, and the IQA and the noise level of three kinds of denoising are increased with the increase of σ, which can be concluded that the adoption of the secondary image denoising method based on weighted kernel norm improvement can achieve good effect no matter the natural image simulation processing or the processing of the high-noise image in real life.
According to the image data processing method provided by the embodiment of the invention, the noise image matrix is subjected to total variation firstly, the residual image matrix is subjected to self-adaptive wiener filtering, and the first denoising image matrix, the filtered residual image matrix and the weight vector are subjected to secondary denoising treatment according to a first preset rule to obtain a second denoising image matrix, wherein the second denoising image matrix fully utilizes prior information of an image, so that the noise is favorably removed, the edge and the detail of the image are better kept, the structural similarity is kept at a higher level while a high signal-to-noise ratio is obtained, and the visual effect of people is fully met.
Fig. 4 is a functional block diagram of an image data processing apparatus 600 according to an embodiment of the present invention. The image data processing 600 includes a first denoising module 610, a processing module 620, a filtering module 630, and a second denoising module 640.
The first denoising module 610 is configured to perform global variational on the noise image matrix to obtain a first denoised image matrix.
The processing module 620 is configured to obtain a residual image matrix according to the first denoised image matrix and the noise image matrix.
The filtering module 630 is configured to perform adaptive wiener filtering on the residual image matrix to obtain a filtered residual image matrix.
The second denoising module 640 is configured to perform a second denoising process on the first denoised image matrix, the filtered residual image matrix, and the weight vector according to a first preset rule, so as to obtain a second denoised image matrix.
The above modules may be implemented by software codes, and in this case, the modules may be stored in the memory 201 of the server 200. The above modules may also be implemented by hardware, such as an integrated circuit chip.
It should be noted that, in the present specification, the embodiments are all described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments may be referred to each other.
The image data processing apparatus provided by the embodiment of the present invention has the same implementation principle and technical effect as the foregoing method embodiments, and for brief description, reference may be made to the corresponding contents in the foregoing method embodiments for the parts of the embodiment that are not mentioned in the apparatus embodiments.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative, and for example, the flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, the functional modules in the embodiments of the present invention may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes. It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention. It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within 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 (5)
1. A method of image data processing, the method comprising:
carrying out total variation on the noise image matrix to obtain a first denoising image matrix;
acquiring a residual image matrix according to the first denoising image matrix and the noise image matrix;
performing self-adaptive wiener filtering on the residual image matrix to obtain a filtered residual image matrix;
adding the first denoised image matrix and the filtered residual image matrix to form a superposition matrix;
dividing the superposition matrix into a plurality of sub-matrices according to the noise variance value;
respectively acquiring a similar matrix corresponding to each sub-matrix according to a second preset rule;
respectively carrying out singular value decomposition on each similar matrix to obtain a first matrix, a singular value diagonal matrix and a second matrix corresponding to each similar matrix;
multiplying the singular value diagonal matrix and the weight vector corresponding to each similar matrix respectively to obtain a third matrix corresponding to each similar matrix;
multiplying the first matrix, the third matrix and the transposed matrix of the second matrix corresponding to each similar matrix respectively to obtain a second denoising submatrix corresponding to each similar matrix;
and summing all the second denoising submatrices to obtain a second denoising image matrix.
2. The method of claim 1, wherein obtaining a residual image matrix from the first denoised image matrix and the noise image matrix comprises:
and subtracting the matrix obtained by the first denoising image matrix from the noise image matrix to obtain a matrix which is used as the residual image matrix.
3. The method of claim 1, wherein the second predetermined rule is a regularization iterative process.
4. The method of claim 3, wherein the regularization iterative formula of the regularization iterative process isWherein,y(0)y is a superposition matrix formed by superposing the first denoised image matrix and the filtered residual image matrix, δ is an iteration step parameter, and k is 6, 8,9 or 11.
5. The method of claim 1, wherein the weight vector w ═ w1,w2,.....,wi]Each element inWhere c is a constant greater than zero, n is a third parameter, and ε is 10-16Is a fixed value and is used as a reference,σi(Yi) Is YiI ═ 1,2.... times, n, Yi of the first denoised image matrix and the filtered residual image matrix are superposed to form a superposition matrix.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610948033.8A CN106504207B (en) | 2016-10-24 | 2016-10-24 | A kind of image processing method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610948033.8A CN106504207B (en) | 2016-10-24 | 2016-10-24 | A kind of image processing method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106504207A CN106504207A (en) | 2017-03-15 |
| CN106504207B true CN106504207B (en) | 2019-08-02 |
Family
ID=58322902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610948033.8A Active CN106504207B (en) | 2016-10-24 | 2016-10-24 | A kind of image processing method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106504207B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108765330B (en) * | 2018-05-21 | 2022-05-27 | 西北大学 | Image denoising method and device based on global and local prior joint constraint |
| CN110675344B (en) * | 2019-09-24 | 2022-07-05 | 福州大学 | Low-rank denoising method and device based on real color image self-similarity |
| CN111383196B (en) * | 2020-03-13 | 2023-07-28 | 浙江大华技术股份有限公司 | Infrared image stripe eliminating method, infrared detector and storage device |
| BR112023026857A2 (en) * | 2021-07-07 | 2024-03-05 | Qualcomm Inc | CASCADE IMAGE PROCESSING FOR NOISE REDUCTION |
| CN113378415B (en) * | 2021-08-12 | 2021-11-02 | 西南科技大学 | Multimedia data self-adaptive recovery method and device based on local and global constraints |
| CN113850632B (en) * | 2021-11-29 | 2022-03-01 | 平安科技（深圳）有限公司 | User category determination method, device, equipment and storage medium |
| CN115601232B (en) * | 2022-12-14 | 2023-03-10 | 华东交通大学 | Color image decoloring method and system based on singular value decomposition |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103208104A (en) * | 2013-04-16 | 2013-07-17 | 浙江工业大学 | Non-local theory-based image denoising method |
| US9015294B2 (en) * | 2006-10-30 | 2015-04-21 | Dell Products L.P. | System and method for assigning addresses to information handling systems |
-
2016
- 2016-10-24 CN CN201610948033.8A patent/CN106504207B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9015294B2 (en) * | 2006-10-30 | 2015-04-21 | Dell Products L.P. | System and method for assigning addresses to information handling systems |
| CN103208104A (en) * | 2013-04-16 | 2013-07-17 | 浙江工业大学 | Non-local theory-based image denoising method |
Non-Patent Citations (1)
| Title |
|---|
| Weighted Nuclear Norm Minimization with Application to Image Denoising;Shuhang Gu 等;《IEEE Conference on Computer Vision and Pattern Recognition（CVPR）》;20141231;第1-8页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106504207A (en) | 2017-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106504207B (en) | A kind of image processing method | |
| JP6239153B2 (en) | Method for removing noise from images with noise | |
| CN106846268A (en) | A kind of Gaussian pulse mixed image noise remove method | |
| CN112802020B (en) | Infrared dim target detection method based on image inpainting and background estimation | |
| Tang et al. | Bayesian framework with non-local and low-rank constraint for image reconstruction | |
| CN114677300A (en) | Hyperspectral image depth noise reduction method and system based on two-stage learning framework | |
| Raza et al. | High density salt and pepper noise removal through decision based partial trimmed global mean filter | |
| Shen et al. | Local thresholding with adaptive window shrinkage in the contourlet domain for image denoising | |
| Hoyer | Independent component analysis in image denoising | |
| Wen et al. | Learning flipping and rotation invariant sparsifying transforms | |
| Mafi et al. | Survey on mixed impulse and Gaussian denoising filters | |
| Hu et al. | Patch-based low-rank minimization for image denoising | |
| Nejati et al. | Low-rank regularized collaborative filtering for image denoising | |
| Sankaran et al. | Non local image restoration using iterative method | |
| Liu et al. | The validity analysis of the non-local mean filter and a derived novel denoising method | |
| Thanh et al. | Distorted image reconstruction method with trimmed median | |
| Ke et al. | A Video Image Compression Method based on Visually Salient Features. | |
| Saikrishna et al. | Detection and removal of random-valued impulse noise from images using sparse representations | |
| Abedini et al. | Image Denoising Using Sparse Representation and Principal Component Analysis | |
| Devi et al. | Metaheuristic Based Noise Identification and Image Denoising Using Adaptive Block Selection Based Filtering | |
| Arora et al. | Bi-featured image quality assessment with the hierarchical image quality enhancement algorithm | |
| Li et al. | An enhanced image denoising method using method noise | |
| RAMADAN | REDUCTION OF MIXED IMPULSIVE NOISE WITH GAUSSIAN AND SPECKLE NOISE | |
| Elsayed et al. | A Fast Image Denoising Algorithm Based on TLS model and Sparse Representation | |
| Talebi et al. | Global denoising is asymptotically optimal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |