CN111783583B - SAR image speckle suppression method based on non-local mean algorithm - Google Patents
SAR image speckle suppression method based on non-local mean algorithm Download PDFInfo
- Publication number
- CN111783583B CN111783583B CN202010574162.1A CN202010574162A CN111783583B CN 111783583 B CN111783583 B CN 111783583B CN 202010574162 A CN202010574162 A CN 202010574162A CN 111783583 B CN111783583 B CN 111783583B
- Authority
- CN
- China
- Prior art keywords
- pixel
- processed
- image
- search domain
- sar image
- 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
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 23
- 230000001629 suppression Effects 0.000 title claims abstract description 22
- 238000001914 filtration Methods 0.000 claims abstract description 47
- 239000011159 matrix material Substances 0.000 claims abstract description 18
- 238000004364 calculation method Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 4
- 238000004088 simulation Methods 0.000 description 19
- 230000000694 effects Effects 0.000 description 8
- 238000009826 distribution Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 230000014759 maintenance of location Effects 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 239000000654 additive Substances 0.000 description 2
- 230000000996 additive effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Radar Systems Or Details Thereof (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a SAR image speckle suppression method based on a non-local mean algorithm, which comprises the following steps: inputting an image, setting the neighborhood size of an image sub-block, and searching the domain size and the gray level difference threshold; selecting a pixel point in an image, and calculating a self-adaptive binary weight matrix of the pixel point; selecting any pixel point in the search domain, calculating the ratio distance between the neighborhood blocks of two pixels, calculating the pixel weight, and calculating the filtered pixel gray value by using the pixel weight; all pixels in the image are subjected to the steps. The invention utilizes the ratio distance to replace Euclidean distance to measure the similarity of two image blocks; removing pixel points with excessive difference from the central pixel in the adjacent area by adopting a binary weighting matrix; when the Gaussian kernel function filtering parameters are selected, a method for selecting the parameters according to the similarity degree of the pixels in the search domain and the center pixels is studied, and unreasonable setting of the filtering parameters is avoided.
Description
Technical Field
The invention relates to the technical field of radar signal processing, in particular to a SAR image speckle suppression method based on a non-local mean algorithm.
Background
The Synthetic Aperture Radar (SAR) is an active microwave imaging radar, and has higher resolution compared with the traditional radar SAR system, so that the SAR can obtain images with high resolution, and is widely applied to the military and civil fields due to good SAR performance. However, the data acquired by the SAR system cannot directly read the information containing the real target, and the image needs to be interpreted, so that the speed and accuracy of the interpretation work are related to the application of SAR in various fields. Compared with the development technology of the SAR system and the SAR imaging technology which are gradually mature, the interpretation research of SAR images is still to be further explored, and the SAR image interpretation work is always a hot subject in the SAR field.
The SAR original image can acquire the target of interest after SAR image interpretation, and speckle suppression of the SAR image is one of the key steps of the SAR image interpretation preprocessing stage. The noise of the coherent speckle can obscure SAR image details, limit the interpretation work of the SAR image, and cause the interpretation failure of the SAR image when serious.
In the speckle suppression method of the image proposed at the present stage, a non-local mean (NLM) algorithm has a good filtering effect on additive Gaussian noise of an optical image, but for a specific multiplicative noise model of an SAR image, the situation that a similarity measurement result is unstable occurs when the NLM algorithm is used for speckle suppression of the SAR image; in addition, for non-uniform samples severely contaminated with noise, performance degradation may also occur.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention aims to provide a SAR image speckle suppression method based on a non-local mean value algorithm, and provides a weighted non-local mean value (WNLMMRD) algorithm based on a maximum ratio distance on the basis of an NLM algorithm, wherein the algorithm utilizes the ratio distance to replace Euclidean distance to measure the similarity of two image blocks; removing pixel points with excessive difference from the central pixel in the adjacent area by adopting a binary weighting matrix; when the Gaussian kernel function filtering parameters are selected, a method for selecting the parameters according to the similarity degree of the pixels in the search domain and the center pixels is studied, and unreasonable setting of the filtering parameters is avoided.
In order to achieve the above purpose, the present invention is realized by the following technical scheme.
The SAR image speckle suppression method based on the non-local mean algorithm comprises the following steps:
step 1, inputting a SAR image to be processed, and reading a gray level data set of the SAR image to be processed; setting the size of a neighborhood block of the SAR image to be processed as N multiplied by N, setting the size of a search domain omega as M multiplied by M, and setting a gray level difference threshold value as ts; n is less than M;
step 2, selecting a pixel point i in the SAR image to be processed as a pixel to be processed, and removing the pixel point with large gray scale difference from the pixel point in the neighborhood block of the pixel to be processed by adopting a binary weighting function to obtain a self-adaptive binary weight matrix T of the pixel i to be processed i ；
Step 3, for any pixel point j, j epsilon omega in the search domain, according to the self-adaptive binary weight matrix T of the pixel i to be processed i Calculating the ratio distance between the pixel i to be processed and each pixel point in the search domain, and further obtaining a Gaussian kernel function filtering parameter h of the pixel i to be processed i ；
Step 4, according to the ratio distance between the pixel i to be processed and each pixel point in the search domain and the Gaussian kernel filtering parameter h corresponding to the ratio distance i Calculating a normalized coefficient z (i) of the pixel i to be processed; normalizing the pixel points in the search domain by adopting a normalization coefficient z (i) of the pixel i to be processed to obtain a corresponding weight w ij ；
Step 5, adopting the weight w of the pixel in the search domain relative to the pixel i to be processed ij Performing self-adaptive non-local mean value filtering processing on the pixel i to be processed to obtain a pixel gray value x 'after filtering of the pixel i to be processed' i ；
And 6, respectively performing steps 2-5 on all pixel points in the SAR image to be processed to obtain a filtered SAR image, namely completing speckle suppression on the input image.
Further, N is an odd number.
Further, the removing pixel points with large difference from the gray scale in the pixel neighborhood block to be processed by adopting the binary weighting function specifically comprises the following steps: a binary weighting function is used:
wherein x is i Representing the gray value, x, of the pixel i to be processed l A gray value representing one pixel point in the neighborhood centered on the pixel i;
then the adaptive binary weight matrix T of the pixel i to be processed i The expression of (2) is: t (T) i ＝{t il ，l∈N i }
Wherein N is i Representing the neighborhood of the center pixel i.
Further, the adaptive binary weight matrix T according to the pixel i to be processed i Calculating the ratio distance between the pixel i to be processed and each pixel point in the search domain, specifically:
(a) Calculating the ratio distance d' between the neighborhood block of the pixel i to be processed and the neighborhood block of the pixel point j in the search domain ij ：
Wherein T is i An adaptive binary weight matrix representing the pixel i to be processed, defining a function of xi (t) =max (t, 1/t), Y i Representing an image block centered on pixel i, Y j Representing an image block centered on pixel j,/represents a dot division,the Gaussian weighted Euclidean distance is represented, and alpha represents the standard deviation of the Gaussian kernel;
(b) And (3) calculating the ratio distance between the rest pixel points in the search domain and the pixel i to be processed according to the step (a).
Further, the Gaussian kernel function filtering parameter h of the pixel i to be processed i The acquisition process of (1) is as follows: and ordering ratio distances between all pixels in the search domain and the neighborhood blocks of the central pixel according to the sequence from large to small, and taking a value at the middle position, namely a median value, as a Gaussian kernel function filtering parameter of the pixel i to be processed.
Further, in step 4, according to the ratio distance between the pixel i to be processed and each pixel point in the search domain, and the gaussian kernel filter parameter h corresponding to the ratio distance i Calculating a normalized coefficient z (i) of the pixel i to be processed, wherein the calculation formula is as follows:
the weight w of pixel j relative to the center pixel i ij The calculation formula of (2) is as follows:
and weight w ij The following conditions are also satisfied:
further, the weighting w of the pixel in the search domain relative to the pixel i to be processed is adopted ij The pixel i to be processed is subjected to self-adaptive non-local mean value filtering processing, and the calculation formula is as follows:
wherein x' i Represents the gray value of pixel i after adaptive non-local mean filtering, Ω represents the search field, x j To search for the gray value, w, of pixel j in the domain ij Representing pixel j relative to the destinationAnd (5) processing the weight of the pixel i.
Compared with the prior art, the invention has the beneficial effects that:
the invention is an improved algorithm based on NLM algorithm. When the SAR image is subjected to speckle suppression, the maximum ratio distance of two image sub-blocks is calculated to replace Euclidean distance in the traditional NLM algorithm so as to measure the similarity of the image sub-blocks; and a binary weighting matrix is used for eliminating pixel points with larger difference, so that the interference of the pixel points on similarity calculation is reduced; in addition, the fluctuation degree of pixels in the search domain is utilized to reasonably select Gaussian kernel function filtering parameters. Compared with the traditional standard NLM algorithm, the method solves the problems that the filtering effect is unstable and the algorithm robustness is low when the SAR image is subjected to speckle suppression.
Drawings
The invention will now be described in further detail with reference to the drawings and to specific examples.
FIG. 1 is a flow chart of an implementation process of the present invention;
FIG. 2 is a synthetic image in an embodiment of the invention;
FIG. 3 is a graph showing the gray value distribution of pixels in the neighborhood of pixels i, j, k of an image according to an embodiment of the present invention; wherein, (a) is a pixel i neighborhood distribution map, (b) is a pixel j neighborhood distribution map, and (c) is a pixel k neighborhood distribution map;
FIG. 4 is a diagram of search windows, neighborhood windows, and similarity metrics in an embodiment of the present invention; wherein, (a) is a schematic diagram of a search window and a neighborhood window, and (b) is a schematic diagram of similarity measurement;
FIG. 5 is a graph showing the results of simulation experiment 1 in the embodiment of the present invention; the method comprises the steps of (a) obtaining an image synthesized by simulation, (b) obtaining a speckle noise image with an additive apparent number L=4, (c) obtaining a Gamma-Map filtering result image under the simulation image, (D) obtaining an NLM algorithm filtering result image under the simulation image, (e) obtaining an SAR-BM3D filtering result image under the simulation image, and (f) obtaining an algorithm filtering result image provided by the invention under the simulation image;
FIG. 6 is a graph showing the results of simulation experiment 2 in the embodiment of the present invention; the method comprises the steps of (a) obtaining a real SAR image, (b) obtaining a Gamma-Map filtering result image under the real SAR image, (c) obtaining an NLM algorithm filtering result image under the real SAR image, (D) obtaining an SAR-BM3D filtering result image under the real SAR image, and (e) obtaining an algorithm filtering result image provided by the invention under the real SAR image.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to examples, but it will be understood by those skilled in the art that the following examples are only for illustrating the present invention and should not be construed as limiting the scope of the present invention.
Referring to fig. 1, the SAR image speckle suppression method based on the non-local mean algorithm provided by the invention is implemented according to the following steps:
step 1, inputting a SAR image to be processed, and reading a gray level data set of the SAR image to be processed; setting the size of a neighborhood block of the SAR image to be processed as N multiplied by N, setting the size of a search domain omega as M multiplied by M, and setting a gray level difference threshold value as ts; n is less than M;
fig. 2 shows an artificially synthesized image, in which the gray level of the upper half image block is 200 and the gray level of the lower half image block is 50, and multiplicative noise with a view number of 4 is added to the image. In fig. 3, (a), (b) and (c) respectively show the distribution of gray values of pixels i, j and k in the neighborhood, where the pixel i is located at the boundary position of the image, and both the pixel j and the pixel k are located in the search domain of the pixel i. Although the center pixels i and j belong to the same gray scale region, since the neighborhood in which the pixel i is located contains two types of pixels, both the neighborhood block in which the pixel j is located and the neighborhood block in which the pixel k is located have gray scale value distributions greatly different from each other. Therefore, since the pixel points located at the boundary as the pixel i have two different types of pixels in the neighborhood block, the similarity calculation can obtain an inaccurate result by using the gray values of all pixels in the neighborhood block, and the pixel points with larger differences from the central pixel in the neighborhood block should be removed before the calculation and then the similarity calculation should be performed.
Step 2, selecting a pixel point i in the SAR image to be processed as a pixel to be processed, and removing the pixel point with large gray scale difference from the pixel point in the neighborhood block of the pixel to be processed by adopting a binary weighting function to obtain a self-adaptive binary weight matrix T of the pixel i to be processed i ；
Fig. 3 (a) is a schematic diagram of a search window and a neighborhood window of an NLM algorithm, fig. 3 (b) is an image to be denoised, first, the size of a neighborhood image block is set to N, Ω represents a search domain with a window of mxm, in order to locate a pixel to be estimated at the center of the neighborhood image block, N is generally an odd number, and it is assumed that the pixel to be estimated is p, q 1 And q 2 For searching two pixel points in the domain, selecting N multiplied by N image sub-blocks as respective neighborhood blocks by taking the three pixel points as centers respectively, and calculating q 1 And q 2 The Euclidean distance between the neighborhood blocks of the pixel point p to be estimated and the neighborhood blocks of the pixel point p to be estimated, and the corresponding weight values are obtained according to the calculation result. In order to eliminate the pixel points with larger difference from the central pixel in the neighborhood block, a binary weighting function is adopted:
wherein x is i The gray value of the center pixel i is represented, and ts is the threshold value of the gray difference. T (T) i ＝{t il ，l∈N i The two-value weighting matrix of the central pixel i can be seen from the above formula, the weight corresponding to the pixel point in the neighborhood, the gray difference value of which is smaller than ts, is 1, i.e. the pixel point normally participates in the calculation of the similarity, and the neighborhood block N i The weight corresponding to the pixel point with the gray difference value larger than ts in the center pixel i is 0, namely the pixel point does not participate in the calculation of the similarity, so that the pixel point positioned at the boundary can avoid the interference of the pixel point with larger difference in the neighborhood block on the calculation of the similarity.
Step 3, for any pixel point j, j epsilon omega in the search domain, according to the self-adaptive binary weight matrix T of the pixel i to be processed i Calculating the ratio distance between the pixel i to be processed and each pixel point in the search domain, and further obtaining a Gaussian kernel function filtering parameter h of the pixel i to be processed i ；
Specifically, the ratio distance d″ between the neighborhood block of pixel i and the neighborhood block of pixel j ij Can be defined as:
wherein T is i ＝{t il ，l∈N i The binary weighting matrix of the center pixel i defines the function xi (t) =max (t, 1/t).
In order to not increase the computational complexity, the method utilizes the fluctuation degree of pixels in the search domain to select an h value for each central pixel, and a specific calculation formula is as follows:
wherein h is i The gaussian kernel function filter parameter representing the center pixel i, the above equation shows that the maximum ratio distance between all pixels in the search domain and the neighborhood block of the center pixel is calculated, and then the median value is taken as the gaussian kernel function filter parameter of the center pixel.
Step 4, according to the ratio distance between the pixel i to be processed and each pixel point in the search domain and the Gaussian kernel filtering parameter h corresponding to the ratio distance i Calculating a normalized coefficient z (i) of the pixel i to be processed; normalizing the pixel points in the search domain by adopting a normalization coefficient z (i) of the pixel i to be processed to obtain a corresponding weight w ij ；
According to the ratio distance d' between the neighborhood block of pixel i and the neighborhood block of pixel j ij Gaussian kernel filter parameter h i Calculating a normalization coefficient z (i), wherein the expression is as follows:
w ij the weight of pixel j with respect to center pixel i is indicated. The expression calculated is:
weight w ij The following conditions are also satisfied:
step 4: using the weights w of pixels in the search domain relative to the pixel i to be processed ij Performing self-adaptive non-local mean value filtering processing on the pixel i to be processed to obtain a pixel gray value x 'after filtering of the pixel i to be processed' i ；
Specifically, an image x= { X is provided 1 ，x 2 ，...，x n X, where x i For the gray value of pixel i, N i Representing an image sub-block with a size of N multiplied by N and taking a pixel point i as a center, and simply called a neighborhood sub-block or a neighborhood block of the pixel i, and then estimating the pixel i by calculating the weighted average of all pixel points in a search domain, wherein the expression is as follows:
wherein x' i Represents the gray value of pixel i after NLM algorithm filtering, Ω represents the search domain with window M×M, x j To search for the gray value, w, of pixel j in the domain ij The weight of the pixel j with respect to the center pixel i, that is, the weight obtained in the step 4.
And 6, respectively performing steps 2-5 on all pixel points in the SAR image to be processed to obtain a filtered SAR image, namely completing speckle suppression on the input image.
Simulation experiment
In order to prove the effectiveness of the invention, a speckle suppression experiment is performed by adopting an analog image and a real SAR image.
Simulation experiment 1 illustrates the effectiveness of the proposed algorithm for an analog image by three objective indicators, peak signal to noise ratio (PSNR), edge hold index (EPI), and equivalent vision number (ENL).
(1) Simulation parameters: specific numerical values of three indexes of PSNR, EPI and ENL are needed to be obtained through calculation, and the calculation formulas of the three indexes are as follows:
wherein MN is the number of image pixels, X (i, j) represents the pixels of the image X without noise,representing filtered pixel points, X s (i, j) represents noise-contaminated image pixels, μ represents the mean value of the filtered image, and σ represents the standard deviation of the filtered image. If the image is an intensity image, the gamma value is 1; if the image is an amplitude image, Γ takes a value of 4/pi-1.
(2) The simulation content:
the specific values of three indexes of PSNR, EPI and ENL of the simulated image can be obtained in the simulation experiment 1 under the simulation parameters, and are shown in the following table:
table 1 analog image filtering evaluation table
In the above table, the larger the peak signal-to-noise ratio (PSNR), the better the filtering effect; the larger the edge hold index (EPI), the stronger the edge hold capability; the greater the equivalent apparent number (ELN), the higher the smoothness of the image.
Fig. 5 (a) shows a synthesized image, (b) shows a synthesized image to which the speckle noise with the apparent number l=4 is added, and (c) to (f) show the filtering results of the Gamma-Map algorithm, the NLM algorithm, the SAR-BM3D algorithm, and the chapter algorithm, respectively. The comparison shows that the algorithm provided by the invention and the SAR-BM3D (three-dimensional block matching) algorithm have excellent filtering effects, the filtering effects of the NLM algorithm are also superior to that of the Gamma-Map algorithm, and the PSNR, EPI and ENL indexes are compared, so that the algorithm has outstanding speckle suppression capability, can obtain satisfactory results in the aspect of edge retention capability, and has texture retention capability which is only inferior to that of the SAR-BM3D algorithm.
Simulation experiment 2 illustrates the effectiveness of the proposed algorithm on a real SAR image by two objective indexes, namely an Edge Preservation Index (EPI) and an equivalent vision number (ENL).
(1) Simulation parameters: the simulation parameters and formulas are the same as those of the simulation experiment 1.
(2) The simulation content: under the simulation parameters, specific numerical values of two indexes of the EPI and the ENL of the real SAR image can be obtained, as shown in the following table:
table 2 real SAR image filtering evaluation table
Also, in the above table, the larger the edge holding index (EPI), the stronger the edge holding ability; the greater the equivalent apparent number (ELN), the higher the smoothness of the image.
Fig. 6 (a) is a real SAR image, and (b) to (e) are filtering results of the Gamma-Map algorithm, the NLM algorithm, the SAR-BM3D algorithm, and the chapter algorithm, respectively. As can be seen from comparison, each algorithm has certain speckle suppression capability and edge retention capability, but the filtering result of the Gamma-Map algorithm is not ideal; compared with Gamma-Map algorithm, NLM algorithm has good speckle suppression capability, and the filtering effect of the NLM algorithm still has room for improvement; the SAR-BM3D algorithm and the algorithm provided by the invention have excellent filtering effects, and the filtering effect of the algorithm is better. Although the texture detail retaining capability in the algorithm processing result is not as good as that of the SAR-BM3D algorithm, the edge retaining capability is better, which also shows that the interference of the pixel points with larger difference from the central pixel in the neighborhood can be reduced by eliminating the pixel points in the similarity measurement calculation.
While the invention has been described in detail in this specification with reference to the general description and the specific embodiments thereof, it will be apparent to one skilled in the art that modifications and improvements can be made thereto. Accordingly, such modifications or improvements may be made without departing from the spirit of the invention and are intended to be within the scope of the invention as claimed.
Claims (5)
1. The SAR image speckle suppression method based on the non-local mean algorithm is characterized by comprising the following steps of:
step 1, inputting a SAR image to be processed, and reading a gray level data set of the SAR image to be processed; setting the size of a neighborhood block of the SAR image to be processed as N multiplied by N, setting the size of a search domain omega as M multiplied by M, and setting a gray level difference threshold value as ts; n < M;
step 2, selecting a pixel point i in the SAR image to be processed as a pixel to be processed, and removing the pixel point with large gray scale difference from the pixel point in the neighborhood block of the pixel to be processed by adopting a binary weighting function to obtain a self-adaptive binary weight matrix T of the pixel i to be processed i ；
Step 3, for any pixel point j, j epsilon omega in the search domain, according to the self-adaptive binary weight matrix T of the pixel i to be processed i Calculating the ratio distance between the pixel i to be processed and each pixel point in the search domain, and further obtaining a Gaussian kernel function filtering parameter h of the pixel i to be processed i ；
The self-adaptive binary weight matrix T according to the pixel i to be processed i Calculating the ratio distance between the pixel i to be processed and each pixel point in the search domain, specifically:
(a) Calculating the ratio distance d between the neighborhood block of the pixel i to be processed and the neighborhood block of the pixel point j in the search domain i ″ j ：
Wherein,T i an adaptive binary weight matrix representing the pixel i to be processed, defining a function of xi (t) =max (t, 1/t), Y i Representing an image block centered on pixel i, Y j Representing an image block centered on pixel j, ·/represents the dot division,the Gaussian weighted Euclidean distance is represented, and alpha represents the standard deviation of the Gaussian kernel;
(b) Calculating the ratio distance between the rest pixel points in the search domain and the pixel i to be processed according to the step (a);
the Gaussian kernel function filtering parameter h of the pixel i to be processed i The acquisition process of (1) is as follows: ordering ratio distances between all pixels in the search domain and the neighborhood blocks of the central pixel according to the sequence from large to small, and then taking a value at the middle position, namely a median value, as a Gaussian kernel function filtering parameter of the pixel i to be processed;
step 4, according to the ratio distance between the pixel i to be processed and each pixel point in the search domain and the Gaussian kernel filtering parameter h corresponding to the ratio distance i Calculating a normalized coefficient z (i) of the pixel i to be processed; normalizing the pixel points in the search domain by adopting a normalization coefficient z (i) of the pixel i to be processed to obtain a corresponding weight w ij ；
Step 5, adopting the weight w of the pixel in the search domain relative to the pixel i to be processed ij Performing self-adaptive non-local mean value filtering processing on the pixel i to be processed to obtain a pixel gray value x 'after filtering of the pixel i to be processed' i ；
And 6, respectively performing steps 2-5 on all pixel points in the SAR image to be processed to obtain a filtered SAR image, namely completing speckle suppression on the input image.
2. The SAR image speckle reduction method based on the non-local mean algorithm of claim 1, wherein N is an odd number.
3. The SAR image speckle suppression method based on the non-local mean algorithm according to claim 1, wherein the method is characterized in that the pixel points with large difference from the gray scale in the pixel neighborhood block to be processed are removed by adopting a binary weighting function, specifically: a binary weighting function is used:
wherein x is i Representing the gray value, x, of the pixel i to be processed l A gray value representing one pixel point in the neighborhood centered on the pixel i;
then the adaptive binary weight matrix T of the pixel i to be processed i The expression of (2) is: t (T) i ＝{t il ,l∈N i }
Wherein N is i Representing a neighborhood block of the center pixel i.
4. The SAR image speckle suppression method based on non-local mean algorithm as set forth in claim 1, wherein in step 4, according to the ratio distance between the pixel i to be processed and each pixel point in the search domain and the Gaussian kernel filter parameter h corresponding to the ratio distance i Calculating a normalized coefficient z (i) of the pixel i to be processed, wherein the calculation formula is as follows:
the weight w of pixel j relative to the center pixel i ij The calculation formula of (2) is as follows:
and weight w ij The following conditions are satisfied:
wherein d i ' j ' represents the ratio distance between the neighborhood block of the pixel i to be processed and the neighborhood block of the pixel point j in the search domain.
5. The SAR image speckle reduction method of claim 1, wherein the weighting w of the pixels in the search domain relative to the pixel i to be processed is used ij The pixel i to be processed is subjected to self-adaptive non-local mean value filtering processing, and the calculation formula is as follows:
wherein x is i ' represents the gray value of pixel i after adaptive non-local mean filtering, Ω represents the search field, x j To search for the gray value, w, of pixel j in the domain ij Representing the weight of pixel j relative to pixel i to be processed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010574162.1A CN111783583B (en) | 2020-06-22 | 2020-06-22 | SAR image speckle suppression method based on non-local mean algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010574162.1A CN111783583B (en) | 2020-06-22 | 2020-06-22 | SAR image speckle suppression method based on non-local mean algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111783583A CN111783583A (en) | 2020-10-16 |
| CN111783583B true CN111783583B (en) | 2024-02-20 |
Family
ID=72756112
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010574162.1A Active CN111783583B (en) | 2020-06-22 | 2020-06-22 | SAR image speckle suppression method based on non-local mean algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111783583B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112698330B (en) * | 2020-12-14 | 2023-09-22 | 大连海事大学 | Non-local average method of polarized SAR image based on ratio logarithmic product |
| CN112614065A (en) * | 2020-12-22 | 2021-04-06 | 航天信息股份有限公司 | Method and system for filtering image data |
| CN112734666B (en) * | 2020-12-31 | 2023-04-07 | 西安电子科技大学 | SAR image speckle non-local mean suppression method based on similarity value |
| CN113221737B (en) * | 2021-05-11 | 2023-09-05 | 杭州海康威视数字技术股份有限公司 | Material information determining method, device, equipment and storage medium |
| CN117115261B (en) * | 2023-10-17 | 2024-03-19 | 深圳市青虹激光科技有限公司 | Knife wheel cutting positioning method and system based on thin wafer |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004312387A (en) * | 2003-04-07 | 2004-11-04 | Sharp Corp | Image processing method, image processing device, image forming device, image pickup device, and computer program |
| WO2010073251A2 (en) * | 2008-12-25 | 2010-07-01 | Medic Vision - Brain Technologies Ltd. | Denoising medical images |
| CN103871039A (en) * | 2014-03-07 | 2014-06-18 | 西安电子科技大学 | Generation method for difference chart in SAR (Synthetic Aperture Radar) image change detection |
| CN103886563A (en) * | 2014-04-18 | 2014-06-25 | 中南民族大学 | SAR image speckle noise inhibition method based on non-local mean and heterogeneity measurement |
| CN109035152A (en) * | 2018-05-23 | 2018-12-18 | 电子科技大学 | A kind of diameter radar image non-local mean filtering method |
| CN110648290A (en) * | 2019-09-06 | 2020-01-03 | 西安交通大学 | Dual-core non-local mean image denoising method based on SURE parameter optimization |
-
2020
- 2020-06-22 CN CN202010574162.1A patent/CN111783583B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004312387A (en) * | 2003-04-07 | 2004-11-04 | Sharp Corp | Image processing method, image processing device, image forming device, image pickup device, and computer program |
| WO2010073251A2 (en) * | 2008-12-25 | 2010-07-01 | Medic Vision - Brain Technologies Ltd. | Denoising medical images |
| CN103871039A (en) * | 2014-03-07 | 2014-06-18 | 西安电子科技大学 | Generation method for difference chart in SAR (Synthetic Aperture Radar) image change detection |
| CN103886563A (en) * | 2014-04-18 | 2014-06-25 | 中南民族大学 | SAR image speckle noise inhibition method based on non-local mean and heterogeneity measurement |
| CN109035152A (en) * | 2018-05-23 | 2018-12-18 | 电子科技大学 | A kind of diameter radar image non-local mean filtering method |
| CN110648290A (en) * | 2019-09-06 | 2020-01-03 | 西安交通大学 | Dual-core non-local mean image denoising method based on SURE parameter optimization |
Non-Patent Citations (4)
| Title |
|---|
| SAR图像相干斑的非局部平均滤波算法;朱磊;蔡飞飞;王延年;郭林源;;西安交通大学学报(第04期);全文 * |
| 基于sigmoid边缘模型的低对比度图像分割算法研究;丁力;周啸虎;陈宇辰;张子齐;;中国医疗设备(第11期);全文 * |
| 基于梯度特征的双核非局部均值去噪算法;张玉征;杨词慧;林泉;;计算机应用研究(第05期);全文 * |
| 基于自适应非局部均值的SAR图像相干斑抑制;陈世媛;李小将;;系统工程与电子技术(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111783583A (en) | 2020-10-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111783583B (en) | SAR image speckle suppression method based on non-local mean algorithm | |
| CN111681174B (en) | Joint filtering method for inhibiting salt-pepper/Gaussian noise mixing target image | |
| CN109919870B (en) | SAR image speckle suppression method based on BM3D | |
| CN111160229B (en) | SSD network-based video target detection method and device | |
| CN111369605B (en) | Infrared and visible light image registration method and system based on edge features | |
| CN109671029A (en) | Image denoising algorithm based on gamma norm minimum | |
| CN107169962B (en) | Gray level image fast segmentation method based on space density constraint kernel fuzzy clustering | |
| CN113191979B (en) | Non-local mean denoising method for partitioned SAR (synthetic aperture radar) image | |
| CN112801141B (en) | Heterogeneous image matching method based on template matching and twin neural network optimization | |
| CN112598069B (en) | Hyperspectral target tracking method based on feature extraction and weight coefficient parameter updating | |
| CN104200434B (en) | Non-local mean image denoising method based on noise variance estimation | |
| CN112598708B (en) | Hyperspectral target tracking method based on four-feature fusion and weight coefficient | |
| CN112508963B (en) | SAR image segmentation method based on fuzzy C-means clustering | |
| CN110766657A (en) | Laser interference image quality evaluation method | |
| CN109242797B (en) | Image denoising method, system and medium based on homogeneous and heterogeneous region fusion | |
| CN111445435A (en) | No-reference image quality evaluation method based on multi-block wavelet transform | |
| CN113436116A (en) | Night image deblurring method based on multi-standard light stripe selection in bipartite graph | |
| Malik et al. | Contrast enhancement and smoothing of CT images for diagnosis | |
| CN111461999A (en) | SAR image speckle suppression method based on super-pixel similarity measurement | |
| CN112734666B (en) | SAR image speckle non-local mean suppression method based on similarity value | |
| CN117392036A (en) | Low-light image enhancement method based on illumination amplitude | |
| CN114066749B (en) | Phase correlation anti-noise displacement estimation method, device and storage medium | |
| CN112686814B (en) | Affine low-rank based image denoising method | |
| CN113759375B (en) | SAR image non-local mean denoising method based on statistical characteristics | |
| CN113506212B (en) | Improved hyperspectral image super-resolution reconstruction method based on POCS |
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 |