CN109035152A - A kind of diameter radar image non-local mean filtering method - Google Patents
A kind of diameter radar image non-local mean filtering method Download PDFInfo
- Publication number
- CN109035152A CN109035152A CN201810502002.9A CN201810502002A CN109035152A CN 109035152 A CN109035152 A CN 109035152A CN 201810502002 A CN201810502002 A CN 201810502002A CN 109035152 A CN109035152 A CN 109035152A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- coefficient
- variation
- noise
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000001914 filtration Methods 0.000 title claims abstract description 42
- 230000001427 coherent effect Effects 0.000 claims abstract description 29
- 230000015556 catabolic process Effects 0.000 claims abstract description 28
- 238000006731 degradation reaction Methods 0.000 claims abstract description 28
- 230000003044 adaptive effect Effects 0.000 claims abstract description 20
- 238000002310 reflectometry Methods 0.000 claims description 9
- 238000011524 similarity measure Methods 0.000 claims description 7
- 238000005315 distribution function Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 5
- 230000000977 initiatory effect Effects 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 3
- 238000000691 measurement method Methods 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 9
- 230000009466 transformation Effects 0.000 description 4
- 230000007850 degeneration Effects 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000556 factor analysis Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003018 immunoassay Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000005191 phase separation Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000010008 shearing Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention proposes a kind of diameter radar image non-local mean filtering methods, belong to radar image process field.The present invention, which estimates heterogeneity, introduces non-local mean algorithm, improves the weight measurement method of non-local mean algorithm.First, for traditional non-local mean method in the method for calculating the weights weighting such as distance measure use, the weighting weight that coefficient of variation calculates pixel distance measure in image block window is introduced herein, since coefficient of variation can characterize the grain direction of image, therefore image block similarity after introducing coefficient of variation weighting can significantly more efficient capture on texture and direction with the image block of similitude, to avoid the fuzzy of marginality and directionality；Then, a kind of adaptive degradation parameter function based on coefficient of variation is devised, adaptive adjustment degradation parameter can effectively guarantee that heterogeneous area obtains protecting not by excess smoothness.So the present invention while filtering out coherent spot, can more effectively keep texture marginal information.
Description
Technical field
The invention belongs to radar image process field, in particular to a kind of diameter radar image non-local mean filtering
Method.
Background technique
Synthetic aperture radar (Synthetic Aperture radar, SAR) is a kind of active high-resolution imaging biography
Sensor, can round-the-clock, it is round-the-clock be observed, penetrability is strong, is widely used in including military investigation, disaster surveillance and ground
Table covers the fields such as Layer Detection.But since SAR imaging is imaged using echo coherent superposition, cause SAR image can not
Avoid there are coherent speckle noise, these coherent speckle noises reduce the improvement of visual effect of SAR image, to subsequent image segmentation and
Target identification also brings difficulty.Therefore, the Speckle reduction of SAR image has weight to the segmentation of SAR image, classification and identification
The meaning wanted.
The phase separation immunoassay method of SAR image can be divided into airspace filter and transform domain filtering.Airspace filter refers to will be to be filtered
The neighborhood territory pixel of wave pixel is weighted and averaged to obtain filtered gray value, and the representative of such methods has Lee filtering, Frost filter
Wave, Gamma-Map filtering etc., they are all based on what specific coherent spot distributed model proposed, therefore can only be to specific distribution
SAR image have preferable filter effect.Transform domain filtering, which refers to, passes through Fourier transformation or wavelet transformation from time domain for image
The methods of transform to corresponding transform domain, be then filtered in transform domain, last inversion gains the side of time domain two dimensional image
Multiple dimensioned, multidirectional analysis method is utilized in method, such methods, and common method has wavelet transformation, warp wavelet, shearing wave
Transformation etc..These two kinds of methods essentially, are all based on the filtering method of part or half part, without considering image
The structural redundancy information of itself.
In order to make full use of the structural redundancy information of image itself, propose non-local mean (Non-Local Means,
NLM) algorithm, the algorithm utilize the structural redundancy of image, the searching image block similar with target image block in image, and according to
Similitude between image block assigns corresponding weight to pixel, and the estimation of target pixel points is acquired using average weighted method
Value, this method can be very good to keep the texture information of image while filtering out noise, therefore also obtain more and more
Focus of attention, and improved on the basis of NLM algorithm.In the prior art, there are also a methods, in the base of NLM algorithm
On plinth, in conjunction with the thought of Wavelet filtering and Wiener filtering, BM3D filtering method is proposed, this method is divided into basis estimation and most
Estimation is estimated twice eventually, is gathered into three-dimensional array similar piece according to a preliminary estimate, is carried out to three-dimensional array using joint hard -threshold
Filtering, filtered image block back to their original positions, the block estimation being overlapped, then weighted average is obtained just
The image of estimation is walked, final estimation is to be gathered into two three using the image of block similarity respectively noisy image and according to a preliminary estimate
Image is tieed up, joint Wiener filtering is then used, filtered image block is put into home position and obtains overlapping estimation image, weighting
After obtain finally estimating image, achieve good results.But the above method all may be only available for additive white Gaussian noise, and
Cannot be in the SAR image with multiplicative noise, therefore, many scholars grind application of the NLM algorithm in SAR image
Study carefully.Therefore, it is proposed to (probability patch-based, PPB) non-local mean filtering based on image block probability distribution
Method, this process employs the noise profile characteristics of image, and maximal possibility estimation is used when calculating block similarity, and
And estimation can be more and more accurate in an iterative process, this method achieves good filter effect in SAR image；Outside it, separately
A kind of method demonstrates ratio and estimates the distance measure for capableing of more steady characterization SAR image, and proposes what sketch-based user interface was estimated
Homogeneous area's non-local mean filtering method, but this method is only applicable to the SAR image denoising in homogeneous area；On the basis of BM3D
It also proposed a kind of SAR-BM3D method, the method for improving block similarity measurement, and solve the problems, such as wavelet coefficient contraction,
Extraordinary filter effect is achieved in SAR image；There is a method in which estimating the basis of non-local mean filtering in ratio
On, the window weight estimated of adjusting the distance improves, and the ratio for enabling Hongxiao Feng to propose estimates non-local mean filtering
Enough it is generalized to heterogeneous area.But these methods, when calculating distance measure, what is used is all pixel gray level distance in window
Mean value or Gauss weighted value, the grain direction information without considering image, and there is a large amount of lines in SAR image
Directional information is managed, these information should become an important evidence of distance measure.In addition, weight is to rely in NLM algorithm
In the gaussian kernel function of distance measure, wherein degradation parameter has a very important influence to filter effect, however few documents
Degradation parameter is studied, is usually configured based on experience value, the effect of self-adaptive solution is not carried out.
Summary of the invention
To solve the above-mentioned problems, the invention proposes a kind of diameter radar image non-local mean filtering method,
It introduces coefficient of variation (coefficient of variation, CV), constructs a kind of according to each to different of image adaptive variation
The two-dimentional weighting windows of property；In addition, a kind of method for proposing adaptive adjustment degradation parameter based on coefficient of variation, makes image same
There is denoising effect well in matter region, while being able to maintain Edge texture details in heterogeneous area.
A kind of diameter radar image non-local mean filtering method, comprising the following steps:
Step 1, synthetic aperture radar magnitude image, initiation parameter are obtained, wherein the parameter of initialization includes searching window
Size Ds, image window size ds, degradation parameter coefficient gamma and S curve gradient parameter ξ；
Step 2, the coherent spot distribution of described image is calculated, is distributed to obtain the noise of coherent spot based on the coherent spot
Variance；
Step 3, according to the standard deviation of described image and mean value computation image degradation coefficient；
Step 4, according to described image coefficient of variation, adaptive degradation parameter is calculated；
Step 5, it is based on the coefficient of variation, constructs adaptive-filtering window；
Step 6, the non local algorithm estimated based on heterogeneity, obtains the value of pixel.
Further, the step 1 further includes following below scheme:
Pixel i is set as the pixel positioned at picture position i, image block i is that the size centered on pixel i is ds's
Image block, field of search ΩiIt is the image block of Ds for the size centered on pixel i.
Further, the step 2 includes following below scheme:
When image pixel spacing and suitable radar imagery resolution cell, noise power is non-relevant, coherent spot right and wrong
Relevant multiplicative noise, is expressed as
υ=un
Wherein, υ is observation, and u is radar reflectivity factor (RCS), and n is coherent speckle noise, according to the standard deviation of image and
Mean value can obtain the equivalent number of image
Wherein, L is equivalent number, and σ is the standard deviation of image, and μ is the mean value of image, and coherent spot is fully developed, intensity
Conditional distribution function is
Wherein, Γ () is Gamma function, set radar reflectivity factor as 1, obtains fully developed coherent speckle noise and divides
Cloth
The variance that noise intensity is calculated is
The range value of described image is the square root of intensity, and the amplitude conditions that can obtain observation are distributed as
Radar reflectivity factor is set as 1, the distribution function of amplitude noise can be obtained
Noise variance, which is calculated, is
Further, the step 3 includes following below scheme:
Coefficient of variation is
Wherein, CV is the local coefficient of variation of pixel.
Further, the step 4 includes following below scheme:
Based on S curve adaptively to adjust coefficient of variation, S curve is expressed as
Wherein, parameter τ is used to control the offset of S curve, and parameter ξ is used to control the gradient of S curve；
According to the coefficient of variation, for each pixel in image, adaptive degradation parameter is expressed as
Wherein, h ' is adaptive degradation factor, h=γ σ, CVmaxFor the local coefficient of variation of maximum of whole image.
Further, the step 5 includes following below scheme:
According to the Multiplicative noise model in the step 2, and noise is the white Gaussian noise that mean value is unit value, a figure
As the coefficient of variation of block is expressed as
Wherein, Ψ is an image block in observed image υ, if the image block is in homogeneity area, E (uΨ 2)=[E
(uΨ)]2, coefficient of variation can be reduced to
It is unit value according to noise figure, can obtainsImage block coefficient of variation in homogeneity area is approximately
Image noise variance is expressed as according to the anisotropic filter window that image local coefficient of variation constructs
ACV(i, k)=exp (- | CV (i, k)-CV (i) |2)
Wherein, ACV(i, k) is the anisotropic weighting windows centered on i, and CV (i, k) is the image block centered on i
In k-th of pixel coefficient of variation.
Further, the step 6 includes following below scheme:
Step 61, image block similarity measurement is calculated
Sketch-based user interface is estimated to be expressed as with the adaptive-filtering window, similarity measure
Wherein, dCV(i, j) is the similarity measure of the image block centered on pixel i and j, and Ψ is image block, ZCV
It (i) is the normalization factor of coefficient of variation in image block, the expression formula of function Ξ () is Ξ (x)=x+1/x, ui(k) for i
Centered on image block k-th of pixel gray value, ujIt (k) is the ash of k-th of pixel of the image block centered on j
Angle value；
Step 62, pixel weight is calculated
According to NLM algorithm, it is based on gaussian kernel function, calculates the weight between pixel
Wherein,For normalization coefficient；
Step 63, pixel pixel value is calculated
According to NLM algorithm, for pixel i, by its field of search ΩiIn pixel by weight be superimposed, obtain pixel i
Estimated value
Wherein, u (j) is the corresponding gray value of pixel j.
Beneficial effects of the present invention: the present invention, which estimates heterogeneity, introduces non-local mean algorithm, improves non local equal
The weight measurement method of value-based algorithm.Firstly, adding for traditional non-local mean method in weights such as calculating distance measure uses
The method of power, introduces the weighting weight that coefficient of variation calculates pixel distance measure in image block window herein, is due to being deteriorated
Number can characterize the grain direction of image, therefore the image block similarity introduced after coefficient of variation weighting significantly more efficient can be caught
The image block on texture and direction with similitude is caught, to avoid the fuzzy of marginality and directionality；Then, one is devised
Adaptive degradation parameter function of the kind based on coefficient of variation, adaptive adjustment degradation parameter can effectively guarantee heterogeneous area
It obtains protecting not by excess smoothness.So the present invention while filtering out coherent spot, can more effectively keep texture edge to believe
Breath.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Fig. 2 is the flow chart of step 6 in Fig. 1.
Specific embodiment
The embodiment of the present invention is described further with reference to the accompanying drawing.
Referring to Fig. 1, a kind of diameter radar image non-local mean filtering method proposed by the present invention, by following
Step is realized:
Step 1, synthetic aperture radar magnitude image, initiation parameter are obtained, wherein the parameter of initialization includes searching window
Size Ds, image window size ds, degradation parameter coefficient gamma and S curve gradient parameter ξ.
In the present embodiment, the image SAR airborne map of magnitudes of input includes homogeneity area image block A and heterogeneous area's figure in image
As block B, tile size is 30 × 30, is used for image filtering outcome evaluation, and wave band is X wave band, and resolution ratio is 3 meters, and image is big
Small is 256 × 256.The parameter of initialization includes searching window size Ds, image window size ds, degradation parameter coefficient gamma and S curve
Gradient parameter ξ.
In the present embodiment, pixel i is set as the pixel positioned at picture position i, image block i is centered on pixel i
Size be ds image block, field of search ΩiIt is the image block of Ds for the size centered on pixel i.
Step 2, the coherent spot distribution of described image is calculated, is distributed to obtain the noise of coherent spot based on the coherent spot
Variance.
In the present embodiment, step 2 is realized by following below scheme:
When image pixel spacing and suitable radar imagery resolution cell, noise power is non-relevant, it is believed that phase
Dry spot is noncoherent multiplicative noise, is expressed as
υ=un
Wherein, υ is observation, and u is true radar reflectivity factor (RCS), and n is coherent speckle noise, according to input picture
Standard deviation and mean value can obtain the equivalent number of image, expression formula is
Wherein, L is equivalent number, and σ is the standard deviation of image, and μ is the mean value of image, for regarding number as the SAR image of L,
If coherent spot is fully developed, strength condition distribution function can be expressed as
Wherein, Γ () is Gamma function, set radar reflectivity factor as 1, obtains fully developed coherent speckle noise and divides
Cloth
The variance that noise intensity is calculated is
The range value of SAR image is the square root of intensity, and the amplitude conditions that can obtain observation are distributed as
Radar reflectivity factor is set as 1, the distribution function of amplitude noise can be obtained
Noise variance, which is calculated, is
Step 3, according to the standard deviation of described image and mean value computation image degradation coefficient.
In the present embodiment, step 3 is realized by following below scheme:
Coefficient of variation is the ratio of graphics standard difference and mean value, i.e.,For each of SAR image pixel,
The local coefficient of variation of the location of pixels is exactly the coefficient of variation for taking the local neighborhood image block of the pixel to estimate the pixel,
The local neighborhood size of pixel is consistent with tile size definition, and the image block of ds × ds centered on pixel is taken to be used for
Calculate the coefficient of variation of central pixel point:
Wherein, σΨAnd μΨFor the standard deviation and mean value of pixel local neighborhood.
Step 4, according to described image coefficient of variation, adaptive degradation parameter is calculated.
In the present embodiment, step 4 is realized by following below scheme:
In traditional NLM algorithm, the degradation factor of the exponential function of weight is calculated in whole picture figure according to similarity measure
In be invariable, therefore the selection of degradation factor is particularly significant, if too big, will lead to excess smoothness, is unfavorable for figure
The holding at the texture edge of picture cannot reach satisfied filter effect in homogeneity area, since coefficient of variation can if too small
Homogeneity area and heterogeneous area, present invention combination coefficient of variation are distinguished, proposes that a kind of adaptive degradation factor calculation method makes to degenerate
The factor can adjust automatically, adapt to different zones image filtering requirements.By the definition of coefficient of variation it is found that the variation in homogeneity area
Coefficient is smaller, and close to noise variance, the coefficient of variation in heterogeneous area is much larger than noise variance, divides in conjunction with above-mentioned degradation factor
Analysis, for the present invention using S type curve come adaptive adjustment coefficient of variation, S type curve representation formula is as follows:
Wherein, parameter τ is used to control the offset of S curve, and parameter ξ is used to control the gradient of S curve.
According to coefficient of variation, for each pixel in image, adaptive degradation parameter is expressed as
Wherein, h ' is adaptive degradation factor, and h is as traditional non-local mean, the value of parameter h and making an uproar for image
Sound variance is related, is an empirical value, generally takes h=γ σ, wherein γ value range can obtain preferably in [90,100]
Effect.Since the coefficient of variation in homogeneity area is close to noise variance, the coefficient of variation in heterogeneous area is much larger than noise variance, therefore joins
The symmetric points position of number τ selection S type curve is equal to the value being calculated when noise variance, i.e., in CV=σ, τ+ξ (CV/CVmax)
=0 solution obtains τ=- ξ (CV/CVmax), the gradient that parameter ξ is used to control S type curve obtains parameter ξ by many experiments
Range better effects can be obtained in [20,50], it is to be noted that, the step is the global system that is deteriorated for whole image
Calculating is counted, wherein the maximum coefficient of variation of the overall situation of parameter list diagram picture.CVmaxFor the local coefficient of variation of maximum of whole image,
For coefficient of variation to be normalized.
Step 5, it is based on the coefficient of variation, constructs adaptive-filtering window.
In the present embodiment, step 5 is realized by following below scheme:
It can be reduced to multiplicative noise according to the coherent speckle noise of step 2 SAR image, if noise is mean value
For the white Gaussian noise of unit value, according to the Multiplicative noise model of step B, the coefficient of variation of an image block can be expressed as
Wherein, Ψ is an image block in observed image υ, if the image block is in homogeneity area, E (uΨ 2)=[E
(uΨ)]2, coefficient of variation can be reduced to
Because noise mean value is unit value, can obtainThat is therefore in the figure in homogeneity area
As block coefficient of variation can be approximated to be image noise variance；If the image block is in heterogeneous area, E (uΨ 2) > > [E
(uΨ)]2, therefore image noise variance is much larger than in the image block coefficient of variation in heterogeneous area, that is to say, that coefficient of variation is at edge
Value at texture is much larger than homogeneous region, using this feature of image local coefficient of variation, can construct it is adaptive it is each to
Anisotropic two-dimensional filtering window introduces the essence that anisotropic filter window is conducive to increase similarity measure in distance measure
Degree, can the high pixel of significantly more efficient selection similitude, so as to more efficiently Protect edge information texture object and point
Target；Meanwhile in homogeneity area, because coefficient of variation is similar to noise variance, filter window near flat, therefore can guarantee each
Filtering to the same sex, that is to say, that in the case where not influencing homogeneous region, take full advantage of the progress of Edge texture directional information
Weighting, effectively captures the direction details of image.The anisotropic filter window constructed according to image local coefficient of variation
It can indicate are as follows:
ACV(i, k)=exp (- | CV (i, k)-CV (i) |2)
Wherein, ACV(i, k) is the anisotropic weighting windows centered on i, and CV (i, k) is the image block centered on i
In k-th of pixel coefficient of variation.If i is located on texture, edge pixel can be invested into high weight using coefficient of variation,
And the pixel unrelated with edge direction invests low weight, to achieve the purpose that direction selection, then ACV(i, k) is along image line
Reason to form filter window；If i is located in point target, the coefficient of variation of point target is much larger than surrounding pixel point, therefore can be with
The form of center weighting is obtained, then ACV(i, k) distribution is similar to gaussian filtering window；If i is located at homogeneity area, due to homogeneity
The coefficient of variation in area is all close to noise variance, therefore each pixel almost invests identical weight, then ACV(i, k) is one
The flat filter window being evenly distributed.
Step 6, the non local algorithm estimated based on heterogeneity, obtains the value of pixel.
Referring to Fig. 2, step 6 is realized by following below scheme:
Step 61, image block similarity measurement is calculated
It can simplify according to coherent spot and be modeled as this feature of multiplicative noise, usage rate of the present invention is estimated similar to calculate
Property measurement, in conjunction with step 5 construct adaptive-filtering window, for the image block in the field of search centered on i and j, similitude
Measure representation is
Wherein, dCV(i, j) is the similarity measure of the image block centered on pixel i and j, and Ψ is image block,For the normalization factor of coefficient of variation in image block, function Ξ's ()
Expression formula is Ξ (x)=x+1/x, r (i, j)=ui(k)/ujIt (k) is the ash of k-th of pixel of the image block centered on i and j
Angle value ui(k) and uj(k) ratio between, it can be seen that r (i, j) illustrates that two gray values are closer, r (i, j) closer to 1
Further away from 1, illustrate that two gray value differences are bigger, and function Ξ () is the function incremental again that first successively decreases at [0 ,+∞],
And minimum value is obtained when r (i, j) is equal to 1.
Step 62, pixel weight is calculated
According to NLM algorithm, the weight of pixel is calculated using gaussian kernel function, wherein using above-mentioned image block similarity
Estimate as distance, degeneration factor is finally normalized, pixel using the adaptive degeneration factor calculated in step 4
Weight between point can be expressed as
Wherein,For normalization coefficient, h ' is adaptive degradation factor, for controlling
The rate of decay of exponential function processed.
Step 63, pixel pixel value is calculated
According to NLM algorithm, for pixel i, by its field of search ΩiIn pixel by weight be superimposed, obtain pixel i
Estimated value
Wherein, u (j) is the corresponding gray value of pixel j.
The present invention is estimated as NL to the pixel value of each pixelCV[u](i).It, can while this method filters out coherent spot
Significantly more efficient holding texture marginal information.
Those of ordinary skill in the art will understand that the embodiments described herein, which is to help reader, understands this hair
Bright principle, it should be understood that protection scope of the present invention is not limited to such specific embodiments and embodiments.This field
Those of ordinary skill disclosed the technical disclosures can make according to the present invention and various not depart from the other each of essence of the invention
The specific variations and combinations of kind, these variations and combinations are still within the scope of the present invention.
Claims (7)
1. a kind of diameter radar image non-local mean filtering method, which comprises the following steps:
Step 1, synthetic aperture radar magnitude image, initiation parameter are obtained, wherein the parameter of initialization includes searching window size
Ds, image window size ds, degradation parameter coefficient gamma and S curve gradient parameter ξ；
Step 2, the coherent spot distribution of described image is calculated, is distributed to obtain the noise side of coherent spot based on the coherent spot
Difference；
Step 3, according to the standard deviation of described image and mean value computation image degradation coefficient；
Step 4, according to described image coefficient of variation, adaptive degradation parameter is calculated；
Step 5, it is based on the coefficient of variation, constructs adaptive-filtering window；
Step 6, the non local algorithm estimated based on heterogeneity, obtains the value of pixel.
2. diameter radar image non-local mean filtering method as described in claim 1, which is characterized in that the step
1 further includes following below scheme:
Pixel i is set as the pixel positioned at picture position i, image block i is the image of ds for the size centered on pixel i
Block, field of search ΩiIt is the image block of Ds for the size centered on pixel i.
3. diameter radar image non-local mean filtering method as claimed in claim 2, which is characterized in that the step
2 include following below scheme:
When image pixel spacing and suitable radar imagery resolution cell, noise power be it is non-relevant, coherent spot is incoherent
Multiplicative noise, be expressed as
υ=un
Wherein, υ is observation, and u is radar reflectivity factor (RCS), and n is coherent speckle noise, according to the standard deviation and mean value of image
The equivalent number of image can be obtained
Wherein, L is equivalent number, and σ is the standard deviation of image, and μ is the mean value of image, and coherent spot is fully developed, strength condition
Distribution function is
Wherein, Γ () is Gamma function, set radar reflectivity factor as 1, obtains fully developed coherent speckle noise and is distributed
The variance that noise intensity is calculated is
The range value of described image is the square root of intensity, and the amplitude conditions that can obtain observation are distributed as
Radar reflectivity factor is set as 1, the distribution function of amplitude noise can be obtained
Noise variance, which is calculated, is
4. diameter radar image non-local mean filtering method as claimed in claim 3, which is characterized in that the step
3 include following below scheme:
Coefficient of variation is
Wherein, CV is the local coefficient of variation of pixel.
5. diameter radar image non-local mean filtering method as claimed in claim 4, which is characterized in that the step
4 include following below scheme:
Based on S curve adaptively to adjust coefficient of variation, S curve is expressed as
Wherein, parameter τ is used to control the offset of S curve, and parameter ξ is used to control the gradient of S curve；
According to the coefficient of variation, for each pixel in image, adaptive degradation parameter is expressed as
Wherein, h ' is adaptive degradation factor, h=γ σ, CVmaxFor the local coefficient of variation of maximum of whole image.
6. diameter radar image non-local mean filtering method as claimed in claim 5, which is characterized in that the step
5 include following below scheme:
According to the Multiplicative noise model in the step 2, and noise is the white Gaussian noise that mean value is unit value, an image block
Coefficient of variation be expressed as
Wherein, Ψ is an image block in observed image υ, if the image block is in homogeneity area, E (uΨ 2)=[E (uΨ)]2, become
Poor coefficient can be reduced to
It is unit value according to noise figure, can obtainsImage block coefficient of variation in homogeneity area is approximately image
Noise variance is expressed as according to the anisotropic filter window that image local coefficient of variation constructs
ACV(i, k)=exp (- | CV (i, k)-CV (i) |2)
Wherein, ACV(i, k) is the anisotropic weighting windows centered on i, and CV (i, k) is kth in the image block centered on i
The coefficient of variation of a pixel.
7. diameter radar image non-local mean filtering method as claimed in claim 6, which is characterized in that the step
6 include following below scheme:
Step 61, image block similarity measurement is calculated
Sketch-based user interface is estimated to be expressed as with the adaptive-filtering window, similarity measure
Wherein, dCV(i, j) is the similarity measure of the image block centered on pixel i and j, and Ψ is image block, ZCVIt (i) is figure
As the normalization factor of coefficient of variation in block, the expression formula of function Ξ () is Ξ (x)=x+1/x, ui(k) for centered on i
The gray value of k-th of pixel of image block, ujIt (k) is the gray value of k-th of pixel of the image block centered on j；
Step 62, pixel weight is calculated
According to NLM algorithm, it is based on gaussian kernel function, calculates the weight between pixel
Wherein,For normalization coefficient；
Step 63, pixel pixel value is calculated
According to NLM algorithm, for pixel i, by its field of search ΩiIn pixel by weight be superimposed, obtain estimating for pixel i
Evaluation
Wherein, u (j) is the corresponding gray value of pixel j.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810502002.9A CN109035152B (en) | 2018-05-23 | 2018-05-23 | Non-local mean filtering method for synthetic aperture radar image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810502002.9A CN109035152B (en) | 2018-05-23 | 2018-05-23 | Non-local mean filtering method for synthetic aperture radar image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109035152A true CN109035152A (en) | 2018-12-18 |
| CN109035152B CN109035152B (en) | 2022-03-18 |
Family
ID=64611660
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810502002.9A Active CN109035152B (en) | 2018-05-23 | 2018-05-23 | Non-local mean filtering method for synthetic aperture radar image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109035152B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110490815A (en) * | 2019-07-12 | 2019-11-22 | 西安理工大学 | A kind of region adaptivity denoising method based on division Bregman algorithm |
| CN110570379A (en) * | 2019-09-11 | 2019-12-13 | 重庆大学 | Non-local mean value CT image noise reduction method based on structure tensor |
| CN110852975A (en) * | 2019-11-15 | 2020-02-28 | 大连海事大学 | Method for jointly driving non-local mean value of radar image by using variation coefficient and local entropy |
| CN111447426A (en) * | 2020-05-13 | 2020-07-24 | 中测新图(北京)遥感技术有限责任公司 | Image color correction method and device |
| CN111461999A (en) * | 2020-03-13 | 2020-07-28 | 西安工程大学 | SAR image speckle suppression method based on super-pixel similarity measurement |
| CN111783583A (en) * | 2020-06-22 | 2020-10-16 | 西安电子科技大学 | SAR image speckle suppression method based on non-local mean algorithm |
| CN112098997A (en) * | 2020-09-18 | 2020-12-18 | 欧必翼太赫兹科技(北京)有限公司 | Three-dimensional holographic imaging security inspection radar image foreign matter detection method |
| CN112381727A (en) * | 2020-11-03 | 2021-02-19 | 中航航空电子有限公司 | Image denoising method and device, computer equipment and storage medium |
| CN112750077A (en) * | 2020-12-14 | 2021-05-04 | 中国船舶重工集团公司第七0九研究所 | Parallelized synthetic aperture radar image sub-band splicing effect processing method |
| CN113160084A (en) * | 2021-04-19 | 2021-07-23 | 新疆大学 | Denoising method and device for quantum dot fluorescence image on surface of porous silicon biosensor |
| CN113160213A (en) * | 2021-05-12 | 2021-07-23 | 西南科技大学 | Novel pixel membership estimation method fusing local and non-local information of image |
| CN113759375A (en) * | 2021-09-15 | 2021-12-07 | 云南师范大学 | SAR image non-local mean de-noising method based on statistical characteristics |
| CN115035326A (en) * | 2022-06-09 | 2022-09-09 | 电子科技大学 | Method for accurately matching radar image and optical image |
| CN116703920A (en) * | 2023-08-07 | 2023-09-05 | 济宁市华祥石墨制品有限公司 | Graphite thermal insulation barrel crack detection method based on image processing |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521811A (en) * | 2011-12-09 | 2012-06-27 | 中国人民解放军海军航空工程学院 | Method for reducing speckles of SAR (synthetic aperture radar) images based on anisotropic diffusion and mutual information homogeneity measuring degrees |
| CN103077499A (en) * | 2013-01-09 | 2013-05-01 | 西安电子科技大学 | SAR (Synthetic Aperture Radar) image pre-processing method based on similar block |
| 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 |
| US20140219552A1 (en) * | 2013-02-07 | 2014-08-07 | Mitsubishi Electric Research Laboratories, Inc. | Denoising of Images with Nonstationary Noise |
| CN103983973A (en) * | 2014-05-28 | 2014-08-13 | 西安电子科技大学 | Synthetic aperture radar imaging method based on image sparse domain noise distribution constraint |
| CN104730500A (en) * | 2015-02-04 | 2015-06-24 | 电子科技大学 | Synthetic aperture radar residual range migration correction method |
| CN105184749A (en) * | 2015-09-18 | 2015-12-23 | 中南民族大学 | Heterogeneity pre-correction-based wavelet domain SAR image despeckling method |
-
2018
- 2018-05-23 CN CN201810502002.9A patent/CN109035152B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521811A (en) * | 2011-12-09 | 2012-06-27 | 中国人民解放军海军航空工程学院 | Method for reducing speckles of SAR (synthetic aperture radar) images based on anisotropic diffusion and mutual information homogeneity measuring degrees |
| CN103077499A (en) * | 2013-01-09 | 2013-05-01 | 西安电子科技大学 | SAR (Synthetic Aperture Radar) image pre-processing method based on similar block |
| US20140219552A1 (en) * | 2013-02-07 | 2014-08-07 | Mitsubishi Electric Research Laboratories, Inc. | Denoising of Images with Nonstationary Noise |
| 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 |
| CN103983973A (en) * | 2014-05-28 | 2014-08-13 | 西安电子科技大学 | Synthetic aperture radar imaging method based on image sparse domain noise distribution constraint |
| CN104730500A (en) * | 2015-02-04 | 2015-06-24 | 电子科技大学 | Synthetic aperture radar residual range migration correction method |
| CN105184749A (en) * | 2015-09-18 | 2015-12-23 | 中南民族大学 | Heterogeneity pre-correction-based wavelet domain SAR image despeckling method |
Non-Patent Citations (6)
| Title |
|---|
| ANTONI BUADES等: "A non-local algorithm for image denoising", 《2005 IEEE COMPUTER SOCIETY CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR"05)》 * |
| QIANG ZHANG等: "Learning a Dilated Residual Network for SAR Image Despeckling", 《ARXIV:1709.02898》 * |
| XIAOLI XING等: "Feature-Based Nonlocal Polarimetric SAR Filtering", 《REMOTE SENSING》 * |
| 易子麟等: "基于非局部均值滤波的SAR图像去噪", 《电子与信息学报》 * |
| 李光廷等: "基于自适应 Bilateral滤波的SAR图像相干斑抑制", 《电子与信息学报》 * |
| 陈世媛等: "基于自适应非局部均值的SAR图像相干斑抑制", 《***工程与电子技术》 * |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110490815A (en) * | 2019-07-12 | 2019-11-22 | 西安理工大学 | A kind of region adaptivity denoising method based on division Bregman algorithm |
| CN110570379A (en) * | 2019-09-11 | 2019-12-13 | 重庆大学 | Non-local mean value CT image noise reduction method based on structure tensor |
| CN110570379B (en) * | 2019-09-11 | 2023-03-24 | 重庆大学 | Non-local mean value CT image noise reduction method based on structure tensor |
| CN110852975A (en) * | 2019-11-15 | 2020-02-28 | 大连海事大学 | Method for jointly driving non-local mean value of radar image by using variation coefficient and local entropy |
| CN110852975B (en) * | 2019-11-15 | 2023-10-17 | 大连海事大学 | Method for driving non-local mean value of radar image by combining variation coefficient and local entropy |
| CN111461999A (en) * | 2020-03-13 | 2020-07-28 | 西安工程大学 | SAR image speckle suppression method based on super-pixel similarity measurement |
| CN111461999B (en) * | 2020-03-13 | 2023-02-14 | 西安工程大学 | SAR image speckle suppression method based on super-pixel similarity measurement |
| CN111447426B (en) * | 2020-05-13 | 2021-12-31 | 中测新图(北京)遥感技术有限责任公司 | Image color correction method and device |
| CN111447426A (en) * | 2020-05-13 | 2020-07-24 | 中测新图(北京)遥感技术有限责任公司 | Image color correction method and device |
| CN111783583B (en) * | 2020-06-22 | 2024-02-20 | 西安电子科技大学 | SAR image speckle suppression method based on non-local mean algorithm |
| CN111783583A (en) * | 2020-06-22 | 2020-10-16 | 西安电子科技大学 | SAR image speckle suppression method based on non-local mean algorithm |
| CN112098997A (en) * | 2020-09-18 | 2020-12-18 | 欧必翼太赫兹科技(北京)有限公司 | Three-dimensional holographic imaging security inspection radar image foreign matter detection method |
| CN112381727A (en) * | 2020-11-03 | 2021-02-19 | 中航航空电子有限公司 | Image denoising method and device, computer equipment and storage medium |
| CN112381727B (en) * | 2020-11-03 | 2023-12-15 | 中航航空电子有限公司 | Image denoising method and device, computer equipment and storage medium |
| CN112750077A (en) * | 2020-12-14 | 2021-05-04 | 中国船舶重工集团公司第七0九研究所 | Parallelized synthetic aperture radar image sub-band splicing effect processing method |
| CN113160084B (en) * | 2021-04-19 | 2022-07-01 | 新疆大学 | Denoising method and device for quantum dot fluorescence image on surface of porous silicon biosensor |
| CN113160084A (en) * | 2021-04-19 | 2021-07-23 | 新疆大学 | Denoising method and device for quantum dot fluorescence image on surface of porous silicon biosensor |
| CN113160213A (en) * | 2021-05-12 | 2021-07-23 | 西南科技大学 | Novel pixel membership estimation method fusing local and non-local information of image |
| CN113759375A (en) * | 2021-09-15 | 2021-12-07 | 云南师范大学 | SAR image non-local mean de-noising method based on statistical characteristics |
| CN113759375B (en) * | 2021-09-15 | 2024-02-09 | 云南师范大学 | SAR image non-local mean denoising method based on statistical characteristics |
| CN115035326A (en) * | 2022-06-09 | 2022-09-09 | 电子科技大学 | Method for accurately matching radar image and optical image |
| CN115035326B (en) * | 2022-06-09 | 2024-04-12 | 电子科技大学 | Radar image and optical image accurate matching method |
| CN116703920A (en) * | 2023-08-07 | 2023-09-05 | 济宁市华祥石墨制品有限公司 | Graphite thermal insulation barrel crack detection method based on image processing |
| CN116703920B (en) * | 2023-08-07 | 2023-10-20 | 济宁市华祥石墨制品有限公司 | Graphite thermal insulation barrel crack detection method based on image processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109035152B (en) | 2022-03-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109035152A (en) | A kind of diameter radar image non-local mean filtering method | |
| CN109919870B (en) | SAR image speckle suppression method based on BM3D | |
| CN106251344B (en) | A kind of multiple dimensioned infrared target self-adapting detecting method of view-based access control model receptive field | |
| Deledalle et al. | Exploiting patch similarity for SAR image processing: The nonlocal paradigm | |
| Guo et al. | Speckle filtering of ultrasonic images using a modified non local-based algorithm | |
| CN109740445B (en) | Method for detecting infrared dim target with variable size | |
| Zhan et al. | Nonlocal means method using weight refining for despeckling of ultrasound images | |
| CN103369209A (en) | Video noise reduction device and video noise reduction method | |
| Ni et al. | Despeckling of SAR image using generalized guided filter with Bayesian nonlocal means | |
| CN106682678B (en) | Image corner detection and classification method based on support domain | |
| US20170301094A1 (en) | Image compounding based on image information | |
| Nair et al. | Fuzzy logic-based automatic contrast enhancement of satellite images of ocean | |
| Bonny et al. | Speckle reduction of ultrasound medical images using Bhattacharyya distance in modified non-local mean filter | |
| CN111861905A (en) | SAR image speckle noise suppression method based on Gamma-Lee filtering | |
| CN103871031A (en) | Kernel regression-based SAR image coherent speckle restraining method | |
| CN109658357A (en) | A kind of denoising method towards remote sensing satellite image | |
| Garzelli et al. | Optimizing SAR change detection based on log-ratio features | |
| CN112698330B (en) | Non-local average method of polarized SAR image based on ratio logarithmic product | |
| Storozhilova et al. | Two approaches for noise filtering in 3D medical CT-images | |
| CN112734666B (en) | SAR image speckle non-local mean suppression method based on similarity value | |
| CN114429593A (en) | Infrared small target detection method based on rapid guided filtering and application thereof | |
| Kumar et al. | Despeckling of polycystic ovary ultrasound images by improved total variation method | |
| Chen et al. | Texture Enhancement Method of Oceanic Internal Waves in SAR Images Based on Non-local Mean Filtering and Multi-scale Retinex | |
| Tong et al. | An improved non-local means filter for SAR image despeckle based on heterogeneity measurement | |
| Zhong et al. | Nonlocal-Lee filter for SAR image despeckling based on hybrid patch similarity |
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 |