CN108765322B - Image denoising method and device - Google Patents

Image denoising method and device Download PDF

Info

Publication number
CN108765322B
CN108765322B CN201810468116.6A CN201810468116A CN108765322B CN 108765322 B CN108765322 B CN 108765322B CN 201810468116 A CN201810468116 A CN 201810468116A CN 108765322 B CN108765322 B CN 108765322B
Authority
CN
China
Prior art keywords
image
denoised
structure group
dictionary
group
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
Application number
CN201810468116.6A
Other languages
Chinese (zh)
Other versions
CN108765322A (en
Inventor
刘金华
任桂平
赖鑫生
蒋昌猛
李永明
吴莲发
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dragon Totem Technology Hefei Co ltd
Original Assignee
Shangrao Normal University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shangrao Normal University filed Critical Shangrao Normal University
Priority to CN201810468116.6A priority Critical patent/CN108765322B/en
Publication of CN108765322A publication Critical patent/CN108765322A/en
Application granted granted Critical
Publication of CN108765322B publication Critical patent/CN108765322B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

The embodiment of the invention provides an image denoising method and device, and relates to the technical field of image processing. The method is applied to an electronic device and comprises the following steps: after an image to be denoised containing mixed noise is obtained, filtering the image to be denoised to obtain an initial image; constructing a plurality of structure groups based on the initial image, and performing dictionary construction on each structure group in the plurality of structure groups to obtain a learning dictionary corresponding to each structure group; then establishing a mixed noise removal model based on the learning dictionary corresponding to each structure group; and performing loop iteration calculation on the mixed noise removal model until a preset iteration number is met, and outputting a denoised image. The method is more accurate and improves the calculation efficiency.

Description

Image denoising method and device
Technical Field
The invention relates to the technical field of image processing, in particular to an image denoising method and device.
Background
The image is easily affected by factors such as a sensor, camera shake, a transmission channel and the like in the transmission process, so that the image is easily polluted by various noises, such as Gaussian mixture noise and impulse noise, and the visual quality of the image is reduced. In order to remove mixed noise in the image, prior and background knowledge of the image is needed, and acquisition of the prior knowledge depends on modeling of the image. The accuracy of the existing image denoising method is low.
Disclosure of Invention
The present invention aims to provide an image denoising method and apparatus, so as to improve the above problems. In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
in a first aspect, an embodiment of the present invention provides an image denoising method applied to an electronic device, where the method includes: acquiring an image to be denoised containing mixed noise; filtering the image to be denoised to obtain an initial image; constructing a plurality of structure groups based on the initial image; performing dictionary construction on each structure group in the plurality of structure groups to obtain a learning dictionary corresponding to each structure group; establishing a mixed noise removal model based on the learning dictionary corresponding to each structure group; and performing loop iteration calculation on the mixed noise removal model until a preset iteration number is met, and outputting a denoised image.
In a second aspect, an embodiment of the present invention provides an image denoising device, which is operated in an electronic device, and includes an obtaining unit, a filtering unit, a structure group constructing unit, a dictionary constructing unit, an establishing unit, and an iterative computation unit. And the acquisition unit is used for acquiring the image to be denoised containing the mixed noise. And the filtering unit is used for carrying out filtering processing on the image to be denoised to obtain an initial image. A structure group construction unit for constructing a plurality of structure groups based on the initial image. And the dictionary construction unit is used for carrying out dictionary construction on each structure group in the plurality of structure groups to obtain a learning dictionary corresponding to each structure group. And the establishing unit is used for establishing a mixed noise removal model based on the learning dictionary corresponding to each structure group. And the iterative computation unit is used for performing cyclic iterative computation on the mixed noise removal model until a preset iteration number is met and outputting a denoised image.
The embodiment of the invention provides an image denoising method and device, which are applied to electronic equipment, wherein the method comprises the following steps: after an image to be denoised containing mixed noise is obtained, filtering the image to be denoised to obtain an initial image; constructing a plurality of structure groups based on the initial image, and performing dictionary construction on each structure group in the plurality of structure groups to obtain a learning dictionary corresponding to each structure group; then establishing a mixed noise removal model based on the learning dictionary corresponding to each structure group; and performing loop iteration calculation on the mixed noise removal model until a preset iteration number is met, and outputting a denoised image. The method is more accurate and improves the calculation efficiency.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the embodiments of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
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 an electronic device applicable to an embodiment of the present invention;
FIG. 2 is a flowchart of an image denoising method according to an embodiment of the present invention;
fig. 3 is a schematic diagram illustrating a comparison of denoising results of Lena images in the image denoising method according to the embodiment of the present invention;
fig. 4 is a schematic diagram illustrating a comparison of denoising results of a boot image in the image denoising method according to the embodiment of the present invention;
fig. 5 is a block diagram of an image denoising device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, 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 some, but not all, embodiments of the present invention. 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 given herein 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.
The inventor of the present application finds the prior art in the examples of the present application: the traditional image mixing denoising scheme of traditional sparse representation and weighted coding is briefly described as follows: firstly, inputting a degraded image y containing Gaussian and salt-pepper mixed noise pollution, setting a dictionary as phi, and carrying out self-adaptive median filtering on the degraded image y to obtain an initial imageImage x(0)Then, the coded residual e is initialized to y-x(0)And initializing mu as 0, wherein the mu represents a non-local similar image block coding coefficient vector. Secondly, joint modeling is carried out on Gaussian mixture and salt and pepper noise, and finally, a model for removing the Gaussian mixture and the salt and pepper noise can be expressed as follows:
Figure BDA0001662566320000041
where w represents the weighting matrix, the diagonal elements wii＝wi。wiiIs shown as
Figure BDA0001662566320000042
In the range of wii∈[0,1]. a is a constant greater than 0 for suppressing the weight wiiThe degree of attenuation of. λ represents a regularization factor, | · | | non-woven phosphor1Is represented by1And (4) norm.
Figure BDA0001662566320000043
Representing the estimated coding coefficient vector. And finally, performing K times of cyclic iterative operation based on the model (1), and finally obtaining the de-noised image x ═ phi alpha(K)。
Further, the inventor of the present application finds that the image mixing denoising method based on the traditional sparse representation and weighted coding analyzes the peak heavy tail distribution characteristics of the mixed noise, and draws a conclusion that the distribution is difficult to model by using the conventional parametric model. The traditional sparse representation and the non-local self-similarity prior information are combined, the coding residual is subjected to weighted modeling to inhibit mixed noise, and a good mixed noise removing effect is achieved. However, in the conventional method, the coded residual is only obtained by the difference between the original noise image and the average value of all pixel points, and when the prior of the non-local self-similar image block is calculated, the conventional method is still used, that is, impulse noise is not eliminated, so that the obtained prior is not accurate enough. On the other hand, each image block needs to be processed independently in the process of dictionary learning and coding by the image mixing denoising method based on traditional sparse representation and weighted coding, so that a high-complexity large-scale optimization problem needs to be solved, the solving of sparse coding coefficients is not accurate, the calculation efficiency of the scheme is low, and the real-time processing requirement is difficult to meet.
In view of this, the embodiments of the present application provide an image denoising method and apparatus, so as to improve accuracy and computational efficiency.
Fig. 1 shows a block diagram of an electronic device 100 applicable to an embodiment of the present invention. As shown in FIG. 1, electronic device 100 may include a memory 102, a memory controller 104, one or more processors 106 (only one shown in FIG. 1), a peripherals interface 108, an input-output module 110, an audio module 112, a display module 114, a radio frequency module 116, and an image denoising mechanism.
The memory 102, the memory controller 104, the processor 106, the peripheral interface 108, the input/output module 110, the audio module 112, the display module 114, and the radio frequency module 116 are electrically connected directly or indirectly to realize data transmission or interaction. For example, electrical connections between these components may be made through one or more communication or signal buses. The image denoising method includes at least one software functional module which can be stored in the memory 102 in the form of software or firmware (firmware), for example, a software functional module or a computer program included in the image denoising apparatus.
The memory 102 may store various software programs and modules, such as program instructions/modules corresponding to the image denoising method and apparatus provided in the embodiments of the present application. The processor 106 executes software programs and modules stored in the memory 102 to execute various functional applications and data processing, that is, to implement the image denoising method in the embodiment of the present application.
The Memory 102 may include, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read Only Memory (PROM), Erasable Read Only Memory (EPROM), electrically Erasable Read Only Memory (EEPROM), and the like.
The processor 106 may be an integrated circuit chip having signal processing capabilities. The Processor may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; but may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components. Which may implement or perform the methods, steps, and logic blocks disclosed in the embodiments of the present application. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The peripherals interface 108 couples various input/output devices to the processor 106 and to the memory 102. In some embodiments, the peripheral interface 108, the processor 106, and the memory controller 104 may be implemented in a single chip. In other examples, they may be implemented separately from the individual chips.
The input-output module 110 is used for providing input data to a user to enable the user to interact with the electronic device 100. The input/output module 110 may be, but is not limited to, a mouse, a keyboard, and the like.
Audio module 112 provides an audio interface to a user that may include one or more microphones, one or more speakers, and audio circuitry.
The display module 114 provides an interactive interface (e.g., a user interface) between the electronic device 100 and a user or for displaying image data to a user reference. In this embodiment, the display module 114 may be a liquid crystal display or a touch display. In the case of a touch display, the display can be a capacitive touch screen or a resistive touch screen, which supports single-point and multi-point touch operations. Supporting single-point and multi-point touch operations means that the touch display can sense touch operations from one or more locations on the touch display at the same time, and the sensed touch operations are sent to the processor 106 for calculation and processing.
The rf module 116 is used for receiving and transmitting electromagnetic waves, and implementing interconversion between the electromagnetic waves and electrical signals, so as to communicate with a communication network or other devices.
It will be appreciated that the configuration shown in FIG. 1 is merely illustrative and that electronic device 100 may include more or fewer components than shown in FIG. 1 or 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 the embodiment of the invention, the electronic device 100 may be a user terminal or a server. The user terminal may be a pc (personal computer), a tablet computer, a mobile phone, a notebook computer, an intelligent television, a set-top box, a vehicle-mounted terminal, and other terminal devices.
In this embodiment, the electronic device is installed with an environment of 64-bit Windows7 os and Matlab 2010. The hardware configuration of the electronic device includes a CPU 2.6G and a memory 8G.
Referring to fig. 2, an embodiment of the present invention provides an image denoising method applied to an electronic device, the method including: step S200, step S210, step S220, step S230, step S240, and step S250.
Step S200: and acquiring an image to be denoised containing mixed noise.
In the present embodiment, the mixed noise may include gaussian noise and salt and pepper noise. For example, the image y to be denoised may be a degraded image contaminated with mixed noise having a gaussian noise variance of 20 and a salt-pepper noise ratio of 30%.
Step S210: and carrying out filtering processing on the image to be denoised to obtain an initial image.
Optionally, step S210 may include:
and carrying out filtering processing on the image to be denoised based on a self-adaptive median filtering algorithm to obtain an initial image.
For example, the image y to be denoised is filtered by the adaptive median filtering algorithm to obtain the initial image x(0)Then, the coded residual e is initialized to y-x(0)。
After step S200, before step S210, the method further comprises:
relevant parameters are initialized.
For example, the relevant parameters may comprise row values B of the image block matrixsThe number of similar blocks c, the regularization factors μ, λ.
Step S220: based on the initial image, a plurality of structure groups are constructed.
Optionally, step S220 may include:
dividing the initial image into a plurality of image blocks; respectively searching adjacent blocks of each image block in the plurality of image blocks based on a block matching algorithm, and obtaining a plurality of similar image blocks corresponding to each image block according to a search result; and forming a structure group by the plurality of similar image blocks corresponding to each image block to obtain a plurality of structure groups.
For example, initial image x(0)Is of size N × N, the initial image is divided into a plurality of sizes Bs×BsImage block xkSuch as BsThe value of (1), (2), n. Then, for each image block, adopting a block matching algorithm to search c similar image blocks which are most matched and similar to the image block, and forming a set by the c similar image blocks
Figure BDA0001662566320000081
Then assemble the
Figure BDA0001662566320000082
All similar image blocks in (1) are arranged into a matrix to form a structural group, which is recorded as
Figure BDA0001662566320000083
Similarly, a plurality of similar image blocks corresponding to each image block are obtained to form a structure group, so that a plurality of structure groups are obtained.
In the embodiment of the application, non-local similarity and sparse prior of an image are combined, image blocks with similarity are searched and form a structure group, a single structure group is used as a basic unit of a sparse representation model instead of a single image block, correlation among the image blocks is considered, and accuracy is improved.
Step S230: and performing dictionary construction on each structure group in the plurality of structure groups to obtain a learning dictionary corresponding to each structure group.
Alternatively, step S230 may include:
and performing dictionary construction on each structure group in the plurality of structure groups based on a singular value decomposition method to obtain a learning dictionary corresponding to each structure group.
For example, for each structural group
Figure BDA0001662566320000084
Corresponding learning dictionary constructed by singular value decomposition method
Figure BDA0001662566320000085
In this embodiment, each structure group has a corresponding dictionary, which has better image content adaptivity, improves accuracy, and improves calculation efficiency.
Step S240: and establishing a mixed noise removal model based on the learning dictionary corresponding to each structure group.
Step S240 may include:
the establishment of the mixed noise removal model comprises the following steps:
Figure BDA0001662566320000091
wherein the content of the first and second substances,
Figure BDA0001662566320000092
for encoding an estimate of a coefficient vector, wGIn order to form a set of weighting matrices,
Figure BDA0001662566320000093
for the ith structure group weight matrix,
Figure BDA0001662566320000094
is shown as
Figure BDA0001662566320000095
Figure BDA0001662566320000096
a is a constant larger than 0, y is the image to be denoised,
Figure BDA0001662566320000097
a learning dictionary, phi, corresponding to the ith structure group of the plurality of structure groupsGFor all that is
Figure BDA0001662566320000098
A combination of (A)GA sparse coding coefficient matrix, alpha, corresponding to the ith structure group of the plurality of structure groupsGFor all that is
Figure BDA0001662566320000099
λ represents a regularization factor, | αG||0Denotes alphaGL of0And (4) norm.
In the embodiment of the application, the mixed noise removal model is a group sparse representation model so as to solve the problem of the computational efficiency of the image mixed noise removal method based on the traditional sparse representation model and the weighted coding.
Step S250: and performing loop iteration calculation on the mixed noise removal model until a preset iteration number is met, and outputting a denoised image.
As an embodiment, step S250 may include:
calling a separation Bregman iterative algorithm to solve and calculate the mixed noise removal model
Figure BDA00016625663200000910
And outputting the denoised image until the preset iteration times are met.
Further, the preset iteration number is K, and the hybrid noise removal model is solved and calculated by calling and separating Bregman iteration algorithm
Figure BDA0001662566320000101
Outputting the denoised image until the preset iteration number is met, wherein the steps comprise:
definition of
Figure BDA0001662566320000102
g(αG)＝λ||αG||0And u is phiGαG；
Calling a separate Bregman iterative algorithm to
Figure BDA0001662566320000103
Converting into the following three formulas to be solved respectively:
Figure BDA0001662566320000104
and
Figure BDA0001662566320000105
μ, λ each represent a regularization factor for 0;
split Bregman iterative algorithm (SBI) solution
Figure BDA0001662566320000106
The optimization problem process is as follows:
given alphaGTo aim at
Figure BDA0001662566320000107
Partial derivative is calculated for u, and an estimated value of u is obtained through calculation
Figure BDA0001662566320000108
Figure BDA0001662566320000109
I represents an identity matrix;
given u, to
Figure BDA00016625663200001010
For alphaGCalculating partial derivative to obtain alphaGIs estimated value of
Figure BDA00016625663200001011
Figure BDA00016625663200001012
Τ＝(λ×BsXc)/(μ N) represents a threshold value, N represents the number of rows or columns of the image matrix,
Figure BDA00016625663200001013
Figure BDA00016625663200001014
representation learning dictionary
Figure BDA00016625663200001015
The inverse matrix of (Hard) · represents a hard threshold operator, <' > represents a Hadamard product operation of two vectors, and 1 represents a vector with all elements being 1;
based on
Figure BDA00016625663200001016
Calculating u(k+1)；
Calculating r(k+1)＝u(k+1)-b(k+1)，Τ＝(λ×Bs×c)/(μN)；
Computing
Figure BDA0001662566320000111
Updating residual e(k)＝y-x(k)According to
Figure BDA0001662566320000112
Calculating structural group weights wG；
All the learning dictionaries
Figure BDA0001662566320000113
Aggregate to update dictionary phiG；
Aggregating all sparse coding coefficients to update
Figure BDA0001662566320000114
Based on
Figure BDA0001662566320000115
Update b(k+1)；
Based on the above, after performing the K-times loop iteration operation, the denoised image x ═ phi is outputGαG (K). Namely, the steps S230 to S240 are repeated, and after K times of loop iteration operation is performed, the denoised image is output.
In order to further explain the denoising performance of the image denoising method provided by the embodiment of the application, two images to be denoised are selected, namely Lena and Boat, and the pixel size is 512 × 512. Meanwhile, the Image Denoising method provided by the embodiment of the application is compared with 2 mixed noise filtering methods in the prior art, wherein the 2 mixed noise filtering methods in the prior art comprise the parameters set according to the original parameter setting in the documents 1, J.Jiang, L.Zhang, J.Yang.Mixed noise removal by weighted encoding with spaced noise regulation, IEEE Transactions on Image Processing,2014,23(6): 2651-derived 2662 and 2, K.Dabov, A.Foi, V.Katkovnik, and K.Egiazarian. Image removal by space 3-D Transform-derived from D-derived video Processing, 2095 (8):2080 and 2092.
In the image denoising method provided by the embodiment of the application, the parameters are set as follows: b issIs 8, i.e. the image block size Bs×BsIs 88; the number c of similar image blocks included in each structure group is set to be 60; the size of the structural group is 64 × 60; the interval between similar image blocks is set to be 4, and the size of a structure group search window is 40 multiplied by 40; the value of the regularization factor lambda is 0.5; the value of μ is set to 2.5e-3。
Referring to fig. 3 and fig. 4, fig. 3 to fig. 4 respectively show the comparison results of the denoising effects of two images to be denoised. Wherein, in fig. 3, (a) shows Lena images polluted by mixed noise, gaussian noise variance is 20, impulse noise density is 30%, fig. 3, (b) shows Lena images denoised by the denoising method of document 2, fig. 3, (c) shows Lena images denoised by the denoising method of document 1, and fig. 3, (d) shows Lena images denoised by the image denoising method provided by the embodiment of the present application; fig. 4 (a) shows a Boat image contaminated by mixed noise, the gaussian noise variance is 20, the impulse noise density is 30%, fig. 4(b) shows the Boat image denoised by the denoising method of document 2, fig. 4(c) shows the Boat image denoised by the denoising method of document 1, and fig. 4(d) shows the Boat image denoised by the image denoising method provided in the embodiment of the present application.
As can be seen from fig. 3 and 4, the denoised Lena image and the denoised bout image output by the image denoising method provided in the embodiment of the present application both have a good edge preserving effect, and have a clear texture structure without an excessive smoothing phenomenon. The image denoising method provided by the embodiment of the application utilizes the structural group sparse representation and the non-local similarity of the image, and combines a weighted coding method to achieve a better mixed noise filtering effect.
Further, the objective evaluation is performed on the hybrid denoising algorithm, and in this embodiment, a (peak signal-to-noise ratio, PSNR) and a (structural similarity index, FSIM) are used to evaluate the quality of the filtered image. Table 1 gives the Lena and Boat images with gaussian noise variance and salt-and-pepper noise density of 10, 20%, respectively; 20. 30 percent; and comparing the performance of the 3 denoising methods. As can be seen from table 1, the image denoising method provided in the embodiment of the present application improves the peak signal-to-noise ratio and the structural similarity index to different extents, and has a good effect of removing gaussian salt and pepper mixed noise.
TABLE 1 comparison of denoised Performance results for Lena and Boat images
Figure BDA0001662566320000131
The embodiment of the invention provides an image denoising method, which is applied to electronic equipment and comprises the following steps: after an image to be denoised containing mixed noise is obtained, filtering the image to be denoised to obtain an initial image; constructing a plurality of structure groups based on the initial image, and performing dictionary construction on each structure group in the plurality of structure groups to obtain a learning dictionary corresponding to each structure group; then establishing a mixed noise removal model based on the learning dictionary corresponding to each structure group; and performing loop iteration calculation on the mixed noise removal model until a preset iteration number is met, and outputting a denoised image. The method is more accurate and improves the calculation efficiency.
Referring to fig. 5, an embodiment of the present invention provides an image denoising apparatus 400, which is operated in an electronic device, where the apparatus 400 includes an obtaining unit 410, a filtering unit 420, a structure group constructing unit 430, a dictionary constructing unit 440, a building unit 450, and an iterative computation unit 460.
An obtaining unit 410, configured to obtain an image to be denoised containing mixed noise.
And the filtering unit 420 is configured to perform filtering processing on the image to be denoised to obtain an initial image.
And the filtering unit 420 is configured to perform filtering processing on the image to be denoised based on an adaptive median filtering algorithm to obtain an initial image.
A structure group construction unit 430, configured to construct a plurality of structure groups based on the initial image.
The structure group constructing unit 430 is configured to: dividing the initial image into a plurality of image blocks; respectively searching adjacent blocks of each image block in the plurality of image blocks based on a block matching algorithm, and obtaining a plurality of similar image blocks corresponding to each image block according to a search result; and forming a structure group by the plurality of similar image blocks corresponding to each image block to obtain a plurality of structure groups.
The dictionary constructing unit 440 is configured to perform dictionary construction on each structure group in the plurality of structure groups, and obtain a learning dictionary corresponding to each structure group.
The dictionary constructing unit 440 is configured to perform dictionary construction on each structure group in the plurality of structure groups based on a singular value decomposition method, so as to obtain a learning dictionary corresponding to each structure group.
The establishing unit 450 is configured to establish a hybrid noise removal model based on the learning dictionary corresponding to each structure group.
The establishing unit 450 is configured to establish a hybrid noise removal model as follows:
Figure BDA0001662566320000141
wherein the content of the first and second substances,
Figure BDA0001662566320000142
for encoding an estimate of a coefficient vector, wGIn order to form a set of weighting matrices,
Figure BDA0001662566320000143
for the ith structure group weight matrix,
Figure BDA0001662566320000144
is shown as
Figure BDA0001662566320000145
a is a constant larger than 0, y is the image to be denoised,
Figure BDA0001662566320000146
a learning dictionary, phi, corresponding to the ith structure group of the plurality of structure groupsGFor all that is
Figure BDA0001662566320000147
A combination of (A)GA sparse coding coefficient matrix, alpha, corresponding to the ith structure group of the plurality of structure groupsGFor all that is
Figure BDA0001662566320000148
λ represents a regularization factor, | αG||0Denotes alphaGL of0And (4) norm.
And the iterative computation unit 460 is configured to perform cyclic iterative computation on the mixed noise removal model until a preset iteration number is met, and output a denoised image.
An iterative computation unit 460 for invoking a separate Bregman iterative algorithm solverComputing the hybrid noise removal model
Figure BDA0001662566320000151
And outputting the denoised image until the preset iteration times are met.
The preset iteration number is K, and the iteration calculating unit 460 is configured to:
definition of
Figure BDA0001662566320000152
g(αG)＝λ||αG||0And u is phiGαG；
Calling a separate Bregman iterative algorithm to
Figure BDA0001662566320000153
Converting into the following three formulas to be solved respectively:
Figure BDA0001662566320000154
and
Figure BDA0001662566320000155
μ, λ each represent a regularization factor for 0;
given alphaGTo aim at
Figure BDA0001662566320000156
Partial derivative is calculated for u, and an estimated value of u is obtained through calculation
Figure BDA0001662566320000157
Figure BDA0001662566320000158
I represents an identity matrix;
given u, to
Figure BDA0001662566320000159
For alphaGCalculating partial derivative to obtain alphaGIs estimated value of
Figure BDA00016625663200001510
Figure BDA00016625663200001511
Τ＝(λ×BsXc)/(μ N) represents a threshold value, N represents the number of rows or columns of the image matrix,
Figure BDA00016625663200001512
Figure BDA00016625663200001513
representation learning dictionary
Figure BDA00016625663200001514
The inverse matrix of (Hard) · represents a hard threshold operator, <' > represents a Hadamard product operation of two vectors, and 1 represents a vector with all elements being 1;
based on
Figure BDA00016625663200001515
Calculating u(k+1)；
Calculating r(k+1)＝u(k+1)-b(k+1)，Τ＝(λ×Bs×c)/(μN)；
Computing
Figure BDA00016625663200001516
Updating residual e(k)＝y-x(k)According to
Figure BDA00016625663200001517
Calculating structural group weights wG；
All the learning dictionaries
Figure BDA0001662566320000161
Aggregate to update dictionary phiG；
Aggregating all sparse coding coefficients to update
Figure BDA0001662566320000162
Based on
Figure BDA0001662566320000163
Update b(k+1)；
Based on the above, after performing the K-times loop iteration operation, the denoised image x ═ phi is outputGαG (K)。
The above units may be implemented by software codes, and in this case, the above units may be stored in the memory 102. The above units may also be implemented by hardware, for example, an integrated circuit chip.
The image denoising apparatus 400 provided in the embodiment of the present invention has the same implementation principle and technical effect as those of the foregoing method embodiments, and for brief description, reference may be made to the corresponding contents in the foregoing method embodiments for what is 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.
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 (9)

1. An image denoising method applied to an electronic device, the method comprising:
acquiring an image to be denoised containing mixed noise;
filtering the image to be denoised to obtain an initial image;
constructing a plurality of structure groups based on the initial image;
performing dictionary construction on each structure group in the plurality of structure groups to obtain a learning dictionary corresponding to each structure group;
establishing a mixed noise removal model based on the learning dictionary corresponding to each structure group;
performing loop iteration calculation on the mixed noise removal model until a preset iteration number is met, and outputting a denoised image;
wherein, the establishing of the mixed noise removal model based on the learning dictionary corresponding to each structure group comprises:
the establishment of the mixed noise removal model comprises the following steps:
Figure FDA0002982762480000011
wherein the content of the first and second substances,
Figure FDA0002982762480000012
for encoding an estimate of a coefficient vector, wGIn order to form a set of weighting matrices,
Figure FDA0002982762480000013
for the ith structure group weight matrix,
Figure FDA0002982762480000014
is shown as
Figure FDA0002982762480000015
Figure FDA0002982762480000016
a is a constant larger than 0, y is the image to be denoised,
Figure FDA0002982762480000017
a learning dictionary, phi, corresponding to the ith structure group of the plurality of structure groupsGFor all that is
Figure FDA0002982762480000018
In the combination of (a) and (b),
Figure FDA0002982762480000019
a sparse coding coefficient matrix, alpha, corresponding to the ith structure group of the plurality of structure groupsGFor all that is
Figure FDA00029827624800000110
λ represents a regularization factor, | αG||0Denotes alphaGL of0And (4) norm.
2. The method of claim 1, wherein constructing a plurality of structural groups based on the initial image comprises:
dividing the initial image into a plurality of image blocks;
respectively searching adjacent blocks of each image block in the plurality of image blocks based on a block matching algorithm, and obtaining a plurality of similar image blocks corresponding to each image block according to a search result;
and forming a structure group by the plurality of similar image blocks corresponding to each image block to obtain a plurality of structure groups.
3. The method according to claim 1, wherein the dictionary construction is performed on each structure group in the plurality of structure groups, and a learning dictionary corresponding to each structure group is obtained, and the method comprises:
and performing dictionary construction on each structure group in the plurality of structure groups based on a singular value decomposition method to obtain a learning dictionary corresponding to each structure group.
4. The method of claim 1, wherein performing a loop iteration calculation on the hybrid noise removal model until a preset number of iterations is satisfied, and outputting a denoised image comprises:
calling a separation Bregman iterative algorithm to solve and calculate the mixed noise removal model
Figure FDA0002982762480000021
And outputting the denoised image until the preset iteration times are met.
5. The method according to claim 4, wherein the preset number of iterations is K, and the hybrid noise removal model is solved and calculated by calling a split Bregman iteration algorithm
Figure FDA0002982762480000022
Outputting the denoised image until the preset iteration number is met, wherein the steps comprise:
definition of
Figure FDA0002982762480000023
g(αG)＝λ||αG||0And u is phiGαG；
Calling a separate Bregman iterative algorithm to
Figure FDA0002982762480000024
Converting into the following three formulas to be solved respectively:
Figure FDA0002982762480000025
Figure FDA0002982762480000031
and
Figure FDA0002982762480000032
μ, λ each represent a regularization factor for 0;
given alphaGTo aim at
Figure FDA0002982762480000033
Partial derivative is calculated for u, and an estimated value of u is obtained through calculation
Figure FDA0002982762480000034
Figure FDA0002982762480000035
I represents an identity matrix;
given u, to
Figure FDA0002982762480000036
For alphaGCalculating partial derivative to obtain alphaGIs estimated value of
Figure FDA0002982762480000037
Figure FDA0002982762480000038
Figure FDA0002982762480000039
T＝(λ×BsX c)/(μ N) represents a threshold value, BsSide length of image block size, c for each junctionThe constellation includes the number of similar image blocks, N represents the number of rows or columns of the image matrix,
Figure FDA00029827624800000310
Figure FDA00029827624800000311
representation learning dictionary
Figure FDA00029827624800000312
The inverse matrix of (Hard) · represents a hard threshold operator, <' > represents a Hadamard product operation of two vectors, and 1 represents a vector with all elements being 1;
based on
Figure FDA00029827624800000313
Calculating u(k+1)；
Calculating r(k+1)＝u(k+1)-b(k+1)，T＝(λ×Bs×c)/(μN)；
Computing
Figure FDA00029827624800000314
Updating residual e(k)＝y-x(k)According to
Figure FDA00029827624800000315
Calculating structural group weights wG；
All the learning dictionaries
Figure FDA00029827624800000316
Aggregate to update dictionary phiG；
Aggregating all sparse coding coefficients to update
Figure FDA00029827624800000317
Based on
Figure FDA00029827624800000318
Update b(k+1)；
Based on the above, after performing the K-times loop iteration operation, the denoised image x ═ phi is outputGαG (K)。
6. The method according to claim 1, wherein the filtering the image to be denoised to obtain an initial image comprises:
and carrying out filtering processing on the image to be denoised based on a self-adaptive median filtering algorithm to obtain an initial image.
7. An image denoising apparatus, operable in an electronic device, the apparatus comprising:
the device comprises an acquisition unit, a denoising unit and a denoising unit, wherein the acquisition unit is used for acquiring an image to be denoised containing mixed noise;
the filtering unit is used for carrying out filtering processing on the image to be denoised to obtain an initial image;
a structure group construction unit for constructing a plurality of structure groups based on the initial image;
the dictionary construction unit is used for carrying out dictionary construction on each structure group in the plurality of structure groups to obtain a learning dictionary corresponding to each structure group;
the establishing unit is used for establishing a mixed noise removing model based on the learning dictionary corresponding to each structure group;
the iterative computation unit is used for performing cyclic iterative computation on the mixed noise removal model until a preset iteration number is met, and outputting a denoised image;
the establishing unit is further configured to establish a hybrid noise removal model as follows:
Figure FDA0002982762480000041
wherein the content of the first and second substances,
Figure FDA0002982762480000042
to a coding systemEstimate of the number vector, wGIn order to form a set of weighting matrices,
Figure FDA0002982762480000043
for the ith structure group weight matrix,
Figure FDA0002982762480000044
is shown as
Figure FDA0002982762480000045
a is a constant larger than 0, y is the image to be denoised,
Figure FDA0002982762480000046
a learning dictionary, phi, corresponding to the ith structure group of the plurality of structure groupsGFor all that is
Figure FDA0002982762480000047
In the combination of (a) and (b),
Figure FDA0002982762480000048
a sparse coding coefficient matrix, alpha, corresponding to the ith structure group of the plurality of structure groupsGFor all that is
Figure FDA0002982762480000049
λ represents a regularization factor, | αG||0Denotes alphaGL of0And (4) norm.
8. The apparatus of claim 7, wherein the structural group construction unit is configured to: dividing the initial image into a plurality of image blocks; respectively searching adjacent blocks of each image block in the plurality of image blocks based on a block matching algorithm, and obtaining a plurality of similar image blocks corresponding to each image block according to a search result; and forming a structure group by the plurality of similar image blocks corresponding to each image block to obtain a plurality of structure groups.
9. The apparatus according to claim 7, wherein the dictionary constructing unit is configured to perform dictionary construction on each structure group in the plurality of structure groups based on a singular value decomposition method, and obtain a learning dictionary corresponding to each structure group.
CN201810468116.6A 2018-05-16 2018-05-16 Image denoising method and device Active CN108765322B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810468116.6A CN108765322B (en) 2018-05-16 2018-05-16 Image denoising method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810468116.6A CN108765322B (en) 2018-05-16 2018-05-16 Image denoising method and device

Publications (2)

Publication Number Publication Date
CN108765322A CN108765322A (en) 2018-11-06
CN108765322B true CN108765322B (en) 2021-04-27

Family

ID=64008201

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810468116.6A Active CN108765322B (en) 2018-05-16 2018-05-16 Image denoising method and device

Country Status (1)

Country Link
CN (1) CN108765322B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109712205A (en) * 2018-12-10 2019-05-03 重庆邮电大学 A kind of compression of images perception method for reconstructing based on non local self similarity model
CN110322407A (en) * 2019-06-03 2019-10-11 辽宁师范大学 Image salt-pepper noise minimizing technology based on depth residual error network
CN111445424B (en) * 2019-07-23 2023-07-18 广州市百果园信息技术有限公司 Image processing method, device, equipment and medium for processing mobile terminal video
CN110728641A (en) * 2019-10-12 2020-01-24 浙江工业大学 Remote sensing image impulse noise removing method and device
CN110661549B (en) * 2019-11-11 2021-05-04 广东石油化工学院 PLC signal reconstruction method and system by utilizing dictionary atoms
CN111242137B (en) * 2020-01-13 2023-05-26 江西理工大学 Spiced salt noise filtering method and device based on morphological component analysis
CN113837958B (en) * 2021-09-09 2023-08-04 南方医科大学 Diffusion weighted image denoising algorithm, medium and device

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105761234A (en) * 2016-01-28 2016-07-13 华南农业大学 Structure sparse representation-based remote sensing image fusion method
CN106157254A (en) * 2015-04-21 2016-11-23 南京理工大学 Rarefaction representation remote sensing images denoising method based on non local self-similarity
CN106204482A (en) * 2016-07-08 2016-12-07 桂林电子科技大学 Based on the mixed noise minimizing technology that weighting is sparse
CN106254720A (en) * 2016-07-19 2016-12-21 四川大学 A kind of video super-resolution method for reconstructing based on associating regularization
WO2017143334A1 (en) * 2016-02-19 2017-08-24 New York University Method and system for multi-talker babble noise reduction using q-factor based signal decomposition
CN107301629A (en) * 2017-06-28 2017-10-27 重庆大学 A kind of image reconstructing method represented based on transform domain joint sparse

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100564592B1 (en) * 2003-12-11 2006-03-28 삼성전자주식회사 Methods for noise removal of moving picture digital data

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106157254A (en) * 2015-04-21 2016-11-23 南京理工大学 Rarefaction representation remote sensing images denoising method based on non local self-similarity
CN105761234A (en) * 2016-01-28 2016-07-13 华南农业大学 Structure sparse representation-based remote sensing image fusion method
WO2017143334A1 (en) * 2016-02-19 2017-08-24 New York University Method and system for multi-talker babble noise reduction using q-factor based signal decomposition
CN106204482A (en) * 2016-07-08 2016-12-07 桂林电子科技大学 Based on the mixed noise minimizing technology that weighting is sparse
CN106254720A (en) * 2016-07-19 2016-12-21 四川大学 A kind of video super-resolution method for reconstructing based on associating regularization
CN107301629A (en) * 2017-06-28 2017-10-27 重庆大学 A kind of image reconstructing method represented based on transform domain joint sparse

Non-Patent Citations (7)

* Cited by examiner, † Cited by third party
Title
"Group-Sparse Representation With Dictionary Learning for Medical Image Denoising and Fusion";Shutao Li et al;《 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING》;20121231;第59卷(第12期);全文 *
"Mixed Noise Removal by Weighted Encoding With Sparse Nonlocal Regularization";Jielin Jiang et al;《IEEE TRANSACTIONS ON IMAGE PROCESSING》;20140630;第23卷(第6期);第2651-2662页 *
"Weighted Joint Sparse Representation for Removing Mixed Noise in Image";Licheng Liu et al;《IEEE TRANSACTIONS ON CYBERNETICS》;20170331;第47卷(第3期);全文 *
"基于稀疏表示模型的图像复原技术研究";张健;《中国博士学位论文全文数据库·信息科技辑》;20141215;第2014年卷(第12期);第1.2-1.3节，第3.4.4节，第四章 *
"基于非局部均值和稀疏表示理论的图像去噪研究";韩玉兰;《中国优秀硕士学位论文全文数据库·信息科技辑》;20151215;第2015年卷(第12期);全文 *
"基于非局部相似性和稀疏表示的图像去噪技术研究";赵井坤;《中国优秀硕士学位论文全文数据库·信息科技辑》;20170315;第2017年卷(第3期);全文 *
张健."基于稀疏表示模型的图像复原技术研究".《中国博士学位论文全文数据库·信息科技辑》.2014,第2014年卷(第12期),第I138-72页. *

Also Published As

Publication number Publication date
CN108765322A (en) 2018-11-06

Similar Documents

Publication Publication Date Title
CN108765322B (en) Image denoising method and device
Zha et al. From rank estimation to rank approximation: Rank residual constraint for image restoration
Yang et al. BM3D-Net: A convolutional neural network for transform-domain collaborative filtering
Rajwade et al. Image denoising using the higher order singular value decomposition
Yang et al. Multitask dictionary learning and sparse representation based single-image super-resolution reconstruction
Wang et al. Image denoising using SVM classification in nonsubsampled contourlet transform domain
CN108805840A (en) Method, apparatus, terminal and the computer readable storage medium of image denoising
CN109766925B (en) Feature fusion method and device, electronic equipment and storage medium
CN113362250B (en) Image denoising method and system based on dual-tree quaternary wavelet and deep learning
Garber et al. Computational models for texture analysis and synthesis
Ono et al. Color-line regularization for color artifact removal
Kim et al. Cauchy noise removal by weighted nuclear norm minimization
Afonso et al. A total variation recursive space-variant filter for image denoising
Chen et al. A soft double regularization approach to parametric blind image deconvolution
Naveed et al. Multiscale image denoising using goodness-of-fit test based on EDF statistics
Kumar et al. Image denoising via overlapping group sparsity using orthogonal moments as similarity measure
Tang et al. Image denoising via graph regularized K-SVD
Sakthidasan Alias Sankaran et al. Noise removal through the exploration of subjective and apparent denoised patches using discrete wavelet transform
Zha et al. Simultaneous nonlocal low-rank and deep priors for poisson denoising
Jon et al. Weighted hyper-Laplacian prior with overlapping group sparsity for image restoration under Cauchy noise
Baljozović et al. Mixed noise removal filter for multi‐channel images based on halfspace deepest location
CN111814884A (en) Target detection network model upgrading method based on deformable convolution
CN116630615A (en) Infrared small target detection method and device, electronic equipment and storage medium
Fan et al. Image denoising by low‐rank approximation with estimation of noise energy distribution in SVD domain
Lv et al. An iterative decoupled method with weighted nuclear norm minimization for image restoration

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
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20231214

Address after: 230000 floor 1, building 2, phase I, e-commerce Park, Jinggang Road, Shushan Economic Development Zone, Hefei City, Anhui Province

Patentee after: Dragon totem Technology (Hefei) Co.,Ltd.

Address before: No. 401, Zhimin Avenue, Xinzhou District, Shangrao City, Jiangxi Province

Patentee before: SHANGRAO NORMAL University