CN108876724A - SAR images filter method based on self-adaptation three-dimensional Block- matching - Google Patents
SAR images filter method based on self-adaptation three-dimensional Block- matching Download PDFInfo
- Publication number
- CN108876724A CN108876724A CN201710324660.9A CN201710324660A CN108876724A CN 108876724 A CN108876724 A CN 108876724A CN 201710324660 A CN201710324660 A CN 201710324660A CN 108876724 A CN108876724 A CN 108876724A
- Authority
- CN
- China
- Prior art keywords
- dimensional
- block
- threshold value
- similar block
- matching
- 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 37
- 239000011159 matrix material Substances 0.000 claims abstract description 75
- 230000009466 transformation Effects 0.000 claims abstract description 32
- 238000001914 filtration Methods 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000001228 spectrum Methods 0.000 claims abstract description 3
- 230000008602 contraction Effects 0.000 claims description 13
- 230000008859 change Effects 0.000 claims description 2
- 230000003044 adaptive effect Effects 0.000 abstract description 3
- 238000005457 optimization Methods 0.000 abstract description 2
- 238000004220 aggregation Methods 0.000 abstract 2
- 230000002776 aggregation Effects 0.000 abstract 2
- 230000000694 effects Effects 0.000 description 4
- 230000010287 polarization Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 206010010904 Convulsion Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/10—Image enhancement or restoration using non-spatial domain filtering
-
- 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
- G06T2207/20004—Adaptive image processing
-
- 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
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to digital image processing techniques fields, and in particular to a kind of SAR images filter method based on self-adaptation three-dimensional Block- matching：SAR image is inputted, image block constructs three-dimensional block matrix；Image blocks Fourier transformation constructs transformation coefficient three-dimensional block matrix；It searches local similar block and constructs similar block three-dimensional matrice；Three-dimensional Fourier transform, three-dimensional inverse transformation and aggregation processing building basis estimation image are carried out to similar block three-dimensional matrice based on adaptive threshold；Basis estimation image block constructs similar block matrix, and corresponding raw video position constructs the similar block matrix of raw video；Adaptive threshold three-dimensional Fourier transform is carried out with raw video three-dimensional similar block matrix to basis estimation；Using basis estimation image as energy spectrum, Wiener filtering optimizes raw video transformation coefficient；Three-dimensional inverse transformation optimization transformation coefficient and aggregation handle to obtain final estimation image.Invention significantly improves the adaptivitys of similar block matrix building efficiency and three-dimension varying filtering.
Description
Technical field
The invention belongs to digital image processing techniques fields, and in particular to a kind of SAR based on self-adaptation three-dimensional Block- matching
Images filter method.
Background technique
Synthetic aperture radar (Synthetic Aperture Radar, SAR) has certain penetrability, can be to vegetation
The region of the ground of covering either desert covering carries out imaging and shows, realizes that target area monitors all-time anf all-weather, leads to
Different images can be obtained in the same area by crossing different polarization modes.Synthetic aperture radar can reach higher resolution ratio,
With very high application value.However SAR image is because there are more speckle noise, the presence of noise for the defect of its image-forming principle
The serious visual identification and subsequent interpretation for hindering SAR image, therefore there is critically important reality to the research of speckle noise
Meaning.
The method being filtered based on image itself similitude can while removing speckle noise preferably reserved graph
The texture information of picture, such as non-local mean filtering are much better than mean filter method.Dabov (2006) is by non local thought and change
It changes domain to combine, BM3D method is applied to image denoising field.BM3D method is by Block- matching some similar piece of structures
Three-dimensional matrice is built up, three-dimension varying is carried out to three-dimensional matrice, coefficient is reconstructed after shrinking by hard -threshold to obtain a base
Then this estimation re-executes Block- matching operation on this basis, denoised with experience Wiener filtering to noise image is originally inputted.
The advantages of partial transformation domain denoising method and non local averaging method is utilized in this method simultaneously, has both introduced less glitch
Image detail is preferably remained again, obtains more satisfactory denoising effect.Subsequent people propose based on wavelet transformation again
The optimization methods such as WBM3D method, the BM1-3D method that three-dimensional bits are filtered based on a peacekeeping three-dimensional space, although these bases
Very high Y-PSNR can be obtained in the algorithm of BM3D, preferable edge is kept to keep index, there is filtering effect well
Fruit, but it still remains following disadvantage：
1) efficiency of algorithm is lower, and similar block lookup is time-consuming, is difficult to handle large scale SAR image；
2) three-dimensional block matrix filtering lacks adaptivity, cannot be according to similar block statistical property automatic adjusument threshold value.
Therefore, in order to solve the above-mentioned problems in the prior art, a kind of new side of self-adaptation three-dimensional Block- matching is needed
Method improves the adaptivity of similar block matrix building efficiency and three-dimension varying filtering.
Summary of the invention
The technical problem to be solved in the invention is：The similar block matrix building of SAR image treatment method in the prior art
Low efficiency, three-dimension varying filtering adaptivity are poor.
It is described that technical scheme is as follows：
A kind of SAR images filter method based on self-adaptation three-dimensional Block- matching, includes the following steps：
Step S1：SAR raw video is inputted, is carried out constructing three-dimensional block matrix according to input image；
Step S2：Fast Fourier is carried out as reference block for each two-dimensional matrix in block matrix three-dimensional in step S1
Transformation, and carry out threshold value contraction；
Step S3：For each reference block after step S2 Fourier transformation, according to reference block Fourier Transform Coefficients
Between l2Norm searches similar block, the three-dimensional similar block matrix of building；
Step S4：Fourier transformation is carried out to the two-dimensional surface signal of three-dimensional similar block matrix, to transformed Fourier
Coefficient is shunk；
Step S5：Fourier transformation is carried out for three-dimensional similar block matrix three-dimensional one-dimensional signal, to transformed Fu
In leaf system number shunk；
Step S6：Three-dimensional inverse Fourier transform obtains the three-dimensional similar block matrix of reconstruct；
Step S7：Whether all reference blocks of three-dimensional block matrix match reconstruct and finish in judgment step 1：It is matched when not completing
Return step S3 when reconstruct carries out matching reconstruct to the reference block for not carrying out matching reconstruct；Enter step when completing matching reconstruct
Rapid S8；
Step S8：The three-dimensional similar block matrix of reconstruct is restored according to the position in raw video, to similar block weight
Folded region is handled using weighted average, obtains basis estimation image；
Step S9：Piecemeal is carried out for basis estimation image, the similar block matrix of basis estimation image three-dimensional is constructed, according to base
Plinth estimates that the similar block matrix of image three-dimensional corresponds to the position in raw video, the three-dimensional similar block matrix of building raw video；
Step S10：Three are carried out with the three-dimensional similar block matrix of raw video for the similar block matrix of basis estimation image three-dimensional
Tie up Fast Fourier Transform (FFT), and carry out threshold value contraction, respectively obtain basis estimation image three-dimensional similar block matrixing coefficient and
Raw video three-dimensional similar block matrixing coefficient；
Step S11：Estimate image three-dimensional similar block matrixing coefficient as true energy in basis after shrinking using threshold value
Raw video three-dimensional similar block matrixing coefficient carries out Wiener filtering after spectrum shrinks threshold value, obtains revised raw video
Three-dimensional similar block matrixing coefficient；
Step S12：It is mutually three-dimensional to revised raw video to carry out three dimensional fast Fourier like block matrix transformation coefficient battle array
Inverse transformation obtains the three-dimensional similar block matrix of reconstruct raw video；
Step S13：Whether judgement basis estimation all reference blocks of image, which match reconstruct, finishes：When not completing matching reconstruct
Return step S9 carries out matching reconstruct to the reference block for not carrying out matching reconstruct；S14 is entered step when completing matching reconstruct；
Step S14：Restore position of the similitude three-dimensional matrice in raw video, weighted average processing similar block overlay region
Domain.
Preferably：
In step S3, in the range of the n × n image block centered on reference block, by reference to block Fourier transformation
L between coefficient2Norm finds similar block, obtains similar block matrix；N can be with value 11.
Preferably：
In step S4 and step S5, transformed Fourier coefficient is shunk using following steps：Initialize hard threshold
Value, the coefficient less than threshold value are set as zero, shrink nonzero term coefficient number in result when threshold value and are more than original non-zero term coefficient
When several M%, increase threshold value, threshold increase steps 1；When nonzero term coefficient number is lower than original nonzero term coefficient number
When N%, reduce threshold value, threshold value reduce step-length be 1, while guarantee threshold value be greater than zero, i.e., when threshold value be less than or equal to zero when, no longer into
Row reduces the operation of threshold value；M can be with value 10 with value 75, N.
Preferably：
Whole non-zero term coefficient in step S8, after being shunk using three-dimensional similar block matrix by three-dimension varying, threshold value
The inverse of number is weighted and averaged as weight；In step S14, using in the three-dimensional similar block coefficient matrix after Wiener filtering
The inverse of nonzero term coefficient number is weighted and averaged as weight.
Beneficial effects of the present invention are：
(1) a kind of SAR images filter method based on self-adaptation three-dimensional Block- matching of the invention, in step S2 and step S3
In, the l after being shunk using threshold value between reference block Fourier Transform Coefficients2Norm carries out solution similar block, after being shunk using threshold value
With higher sparsity, operation efficiency is higher, simplifies the building complexity of similar block matrix, significantly improves similar block matrix
Construct efficiency；
(2) a kind of SAR images filter method based on self-adaptation three-dimensional Block- matching of the invention, in step S4 and step S5
In, hard -threshold is initialized using input picture statistical property, and according to the result of collapse threshold in actual threshold contraction process
Threshold value is gradually adjusted, carries out self-adaptive processing for the image blocks of different statistical properties, is avoided corresponding to three-dimensional similar block matrix
The contraction of Fourier Transform Coefficients is improper to cause image texture detailed information to lose the serious or unconspicuous situation of filter effect；
(3) a kind of SAR images filter method based on self-adaptation three-dimensional Block- matching of the invention, in step s 11, with base
Plinth is estimated to be modified raw video transformation coefficient by Wiener filtering based on similar block conversion coefficient, avoids relying solely on
The pseudo- texture phenomenon of brought image is estimated on basis, expands the filter effect that adaptive threshold generates.
Detailed description of the invention
Fig. 1 is a kind of SAR images filter method flow diagram based on self-adaptation three-dimensional Block- matching of the invention.
Specific embodiment
With reference to the accompanying drawings of the specification with embodiment to a kind of SAR image based on self-adaptation three-dimensional Block- matching of the invention
Filtering method is described in detail.
As shown in Figure 1, a kind of SAR images filter method based on self-adaptation three-dimensional Block- matching of the invention, including it is following
Step：
Step S1
SAR raw video is inputted, is carried out constructing three-dimensional block matrix according to input image.
In the present embodiment, input RADASAT2 satellite C-band HH polarization, the SAR image that size is 1024x1024, setting
Tile size is 8 × 8, is 4 building three-dimensional bits matrix B locks with smooth step-length.
Step S2
Fast Fourier Transform (FFT) is carried out as reference block for each two-dimensional matrix in block matrix three-dimensional in step S1, and
Carry out threshold value contraction.
In the present embodiment, Fourier transformation is carried out to each reference block and carries out threshold value contraction, hard -threshold Th1 is 2.8.
The initialization mode of hard -threshold Th1 is the intermediate value of whole input image Fourier Transform Coefficients.
Step S3
For each reference block after step S2 Fourier transformation, according to the l between reference block Fourier Transform Coefficients2Model
Number searches similar block, the three-dimensional similar block matrix of building.
In the present embodiment, in reference block upper and lower, left and right in the range of each 5 image blocks, i.e., centered on reference block
11 × 11 image blocks in the range of, by reference to the l between block Fourier Transform Coefficients2Norm finds similar block, similar block
Threshold value Th2 is 11.7, and most similar block numbers are 28, obtains similar block matrix.
Step S4
Fourier transformation is carried out to the two-dimensional surface signal of three-dimensional similar block matrix, hard -threshold is initialized, to transformed
Fourier coefficient is shunk, and the coefficient less than threshold value is set as zero, is shunk nonzero term coefficient number in result when threshold value and is more than
Original nonzero term coefficient number 75% when, increase threshold value, threshold increase steps 1；When nonzero term coefficient number is lower than original
Nonzero term coefficient number 10% when, reduce threshold value, threshold value reduce step-length be 1, while guarantee threshold value be greater than zero, i.e., when threshold value is small
When being equal to zero, the operation for reducing threshold value is no longer carried out.
In the present embodiment, hard -threshold is initialized using generic threshold value shown in following formula：
In formula, tau is generic threshold value, and σ is two-dimensional surface signal variance, and N is two-dimensional surface signal length.
Step S5
Fourier transformation is carried out for three-dimensional similar block matrix three-dimensional one-dimensional signal, hard -threshold is initialized, to transformation
Fourier coefficient afterwards is shunk, and the coefficient less than threshold value is set as zero, when threshold value shrinks nonzero term coefficient number in result
More than original nonzero term coefficient number 75% when, increase threshold value, threshold increase steps 1；When nonzero term coefficient number is lower than
Original nonzero term coefficient number 10% when, reduce threshold value, threshold value reduce step-length be 1, while guarantee threshold value be greater than zero, that is, work as threshold
When value is less than or equal to zero, the operation for reducing threshold value is no longer carried out.
It should be noted that when three-dimensional signal number, that is, similar block number is less than 3, or work as input signal non-zero
When term coefficient number is less than the 25% of input signal number, without Fourier transformation, retain three-dimensional signal.
Step S6
Three-dimensional inverse Fourier transform obtains the three-dimensional similar block matrix of reconstruct.
Step S7
Whether all reference blocks of three-dimensional block matrix match reconstruct and finish in judgment step 1：When not completing matching reconstruct
Return step S3 carries out matching reconstruct to the reference block for not carrying out matching reconstruct；S8 is entered step when completing matching reconstruct.
Step S8
The three-dimensional similar block matrix of reconstruct is restored according to the position in raw video, similar block overlapping region is adopted
It is handled with weighted average, obtains basis estimation image.
Whole non-zero term system in the present embodiment, after being shunk using three-dimensional similar block matrix by three-dimension varying, threshold value
The inverse of several numbers is weighted and averaged as weight.
Step S9,
Piecemeal is carried out for basis estimation image, the similar block matrix of basis estimation image three-dimensional is constructed, is estimated according to basis
The similar block matrix of image three-dimensional corresponds to the position in raw video, the three-dimensional similar block matrix of building raw video.
In the present embodiment, piecemeal is carried out to basis estimation image, each reference block image of basis estimation image is carried out
Fast Fourier Transform (FFT) estimates each reference block of image according to the l between Fourier Transform Coefficients for basis2Norm searches phase
Like block, the similar block matrix of basis estimation image three-dimensional is constructed.Wherein, with reference to block size and its threshold value setting with step S3.According to
Position of the estimation image three-dimensional similar block matrix in basis in raw video, the three-dimensional similar block matrix of building raw video.
Step S10
Three-dimensional quick Fu is carried out with the three-dimensional similar block matrix of raw video for the similar block matrix of basis estimation image three-dimensional
In leaf transformation, and carry out threshold value contraction, respectively obtain basis estimation image three-dimensional similar block matrixing coefficient and raw video
Three-dimensional similar block matrixing coefficient.
In the present embodiment, threshold value shrinks mode with step S4 and S5, shrinks, is less than to transformed Fourier coefficient
The coefficient of threshold value is set as zero, shrinks nonzero term coefficient number in result when threshold value and is more than the 75% of original nonzero term coefficient number
When, increase threshold value, threshold increase steps 1；When nonzero term coefficient number is lower than the 10% of original nonzero term coefficient number, subtract
Few threshold value, it is 1 that threshold value, which reduces step-length, while guaranteeing that threshold value is greater than zero.
Step S11
Basis estimation image three-dimensional similar block matrixing coefficient after being shunk using threshold value is composed as true energy to threshold value
Raw video three-dimensional similar block matrixing coefficient carries out Wiener filtering after contraction, and it is similar to obtain revised raw video three-dimensional
Block matrix transformation coefficient.
Step S12
It is mutually three-dimensional to revised raw video to carry out three dimensional fast Fourier inverse transformation like block matrix transformation coefficient battle array, it obtains
To the three-dimensional similar block matrix of reconstruct raw video.
Step S13
Whether judgement basis estimation all reference blocks of image, which match reconstruct, finishes：The return step when not completing matching reconstruct
S9 carries out matching reconstruct to the reference block for not carrying out matching reconstruct；S14 is entered step when completing matching reconstruct.
Step S14
Restore position of the similitude three-dimensional matrice in raw video, weighted average processing similar block overlapping region.
In the present embodiment, the three-dimensional similar block matrix of reconstruct raw video is gone back according to the position in raw video
Original is handled similar block overlapping region using weighted average, i.e., by non-zero in the three-dimensional similar block coefficient matrix after Wiener filtering
As weight, the weighted average for carrying out repeat region obtains finally estimating reconstructed image the inverse of term coefficient number.
Combining most preferred embodiment above, invention has been described, but the invention is not limited to implementations disclosed above
Example, and various modifications, equivalent combinations according to the essence of the present invention should be covered.
Claims (10)
1. a kind of SAR images filter method based on self-adaptation three-dimensional Block- matching, it is characterised in that：Include the following steps：
Step S1：SAR raw video is inputted, is carried out constructing three-dimensional block matrix according to input image；
Step S2：Fast Fourier change is carried out as reference block for each two-dimensional matrix in block matrix three-dimensional in step S1
It changes, and carries out threshold value contraction；
Step S3：For each reference block after step S2 Fourier transformation, according to the l between reference block Fourier Transform Coefficients2
Norm searches similar block, the three-dimensional similar block matrix of building；
Step S4：Fourier transformation is carried out to the two-dimensional surface signal of three-dimensional similar block matrix, to transformed Fourier coefficient
It is shunk；
Step S5：Fourier transformation is carried out for three-dimensional similar block matrix three-dimensional one-dimensional signal, to transformed Fourier
Coefficient is shunk；
Step S6：Three-dimensional inverse Fourier transform obtains the three-dimensional similar block matrix of reconstruct；
Step S7：Whether all reference blocks of three-dimensional block matrix match reconstruct and finish in judgment step 1：It is reconstructed when not completing matching
When return step S3, to do not carry out matching reconstruct reference block carry out matching reconstruct；S8 is entered step when completing matching reconstruct；
Step S8：The three-dimensional similar block matrix of reconstruct is restored according to the position in raw video, to similar block overlay region
Domain is handled using weighted average, obtains basis estimation image；
Step S9：Piecemeal is carried out for basis estimation image, the similar block matrix of basis estimation image three-dimensional is constructed, is estimated according to basis
It counts the similar block matrix of image three-dimensional and corresponds to the position in raw video, the three-dimensional similar block matrix of building raw video；
Step S10：It is three-dimensional fast for the similar block matrix of basis estimation image three-dimensional and the three-dimensional similar block matrix progress of raw video
Fast Fourier transformation, and carry out threshold value contraction respectively obtains basis estimation image three-dimensional similar block matrixing coefficient and original
Image three-dimensional similar block matrixing coefficient；
Step S11：Estimate image three-dimensional similar block matrixing coefficient as true energy spectrum pair in basis after shrinking using threshold value
Raw video three-dimensional similar block matrixing coefficient carries out Wiener filtering after threshold value is shunk, and it is three-dimensional to obtain revised raw video
Similar block matrixing coefficient；
Step S12：It is mutually three-dimensional to revised raw video to carry out three dimensional fast Fourier inversion like block matrix transformation coefficient battle array
It changes, obtains the three-dimensional similar block matrix of reconstruct raw video；
Step S13：Whether judgement basis estimation all reference blocks of image, which match reconstruct, finishes：It is returned when not completing matching reconstruct
Step S9 carries out matching reconstruct to the reference block for not carrying out matching reconstruct；S14 is entered step when completing matching reconstruct；
Step S14：Restore position of the similitude three-dimensional matrice in raw video, weighted average processing similar block overlapping region.
2. the SAR images filter method according to claim 1 based on self-adaptation three-dimensional Block- matching, it is characterised in that：Step
In rapid S3, in the range of the n × n image block centered on reference block, by reference to the l between block Fourier Transform Coefficients2
Norm finds similar block, obtains similar block matrix.
3. the SAR images filter method according to claim 2 based on self-adaptation three-dimensional Block- matching, it is characterised in that：N=
11。
4. the SAR images filter method according to claim 1 or 2 based on self-adaptation three-dimensional Block- matching, it is characterised in that：
In step S4, transformed Fourier coefficient is shunk using following steps：Hard -threshold is initialized, less than the coefficient of threshold value
It is set as zero, when nonzero term coefficient number is more than the M1% of original nonzero term coefficient number in threshold value contraction result, increases threshold
Value, threshold increase steps 1；When nonzero term coefficient number is lower than the N1% of original nonzero term coefficient number, threshold value, threshold are reduced
It is 1 that value, which reduces step-length, while guaranteeing that threshold value is greater than zero, i.e., when threshold value is less than or equal to zero, no longer carries out the operation for reducing threshold value.
5. the SAR images filter method according to claim 4 based on self-adaptation three-dimensional Block- matching, it is characterised in that：M1
=75, N1=10.
6. the SAR images filter method according to claim 1 or 2 based on self-adaptation three-dimensional Block- matching, it is characterised in that：
In step S5, transformed Fourier coefficient is shunk using following steps：Hard -threshold is initialized, less than the coefficient of threshold value
It is set as zero, when nonzero term coefficient number is more than the M2% of original nonzero term coefficient number in threshold value contraction result, increases threshold
Value, threshold increase steps 1；When nonzero term coefficient number is lower than the N2% of original nonzero term coefficient number, threshold value, threshold are reduced
It is 1 that value, which reduces step-length, while guaranteeing that threshold value is greater than zero, i.e., when threshold value is less than or equal to zero, no longer carries out the operation for reducing threshold value.
7. the SAR images filter method according to claim 6 based on self-adaptation three-dimensional Block- matching, it is characterised in that：M2
=75, N2=10.
8. the SAR images filter method according to claim 7 based on self-adaptation three-dimensional Block- matching, it is characterised in that：When
When three-dimensional signal number, that is, similar block number is less than 3, or when input signal nonzero term coefficient number is less than input signal
Number 25% when, without Fourier transformation, retain three-dimensional signal.
9. the SAR images filter method according to claim 1 or 2 based on self-adaptation three-dimensional Block- matching, it is characterised in that：
In step S8, falling for whole nonzero term coefficient number of the three-dimensional similar block matrix after three-dimension varying, threshold value contraction is utilized
Number is weighted and averaged as weight.
10. the SAR images filter method according to claim 1 or 2 based on self-adaptation three-dimensional Block- matching, feature exist
In：In step S14, using the inverse of nonzero term coefficient number in the three-dimensional similar block coefficient matrix after Wiener filtering as weight
It is weighted and averaged.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710324660.9A CN108876724B (en) | 2017-05-10 | 2017-05-10 | SAR image filtering method based on self-adaptive three-dimensional block matching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710324660.9A CN108876724B (en) | 2017-05-10 | 2017-05-10 | SAR image filtering method based on self-adaptive three-dimensional block matching |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108876724A true CN108876724A (en) | 2018-11-23 |
| CN108876724B CN108876724B (en) | 2021-10-22 |
Family
ID=64287941
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710324660.9A Active CN108876724B (en) | 2017-05-10 | 2017-05-10 | SAR image filtering method based on self-adaptive three-dimensional block matching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108876724B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110954921A (en) * | 2019-12-03 | 2020-04-03 | 浙江大学 | Laser radar echo signal-to-noise ratio improving method based on block matching 3D collaborative filtering |
| CN114509749A (en) * | 2022-04-19 | 2022-05-17 | 亿慧云智能科技（深圳）股份有限公司 | Indoor positioning detection system and method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102682429A (en) * | 2012-04-13 | 2012-09-19 | 泰山学院 | De-noising method of filtering images in size adaptive block matching transform domains |
| CN102722879A (en) * | 2012-06-13 | 2012-10-10 | 西安电子科技大学 | SAR (synthetic aperture radar) image despeckle method based on target extraction and three-dimensional block matching denoising |
| US20150206504A1 (en) * | 2014-01-21 | 2015-07-23 | Nvidia Corporation | Unified optimization method for end-to-end camera image processing for translating a sensor captured image to a display image |
| CN105550997A (en) * | 2015-12-08 | 2016-05-04 | 天津津航计算技术研究所 | Three-dimensional matching image denoising method based on multiple transform domains |
-
2017
- 2017-05-10 CN CN201710324660.9A patent/CN108876724B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102682429A (en) * | 2012-04-13 | 2012-09-19 | 泰山学院 | De-noising method of filtering images in size adaptive block matching transform domains |
| CN102722879A (en) * | 2012-06-13 | 2012-10-10 | 西安电子科技大学 | SAR (synthetic aperture radar) image despeckle method based on target extraction and three-dimensional block matching denoising |
| US20150206504A1 (en) * | 2014-01-21 | 2015-07-23 | Nvidia Corporation | Unified optimization method for end-to-end camera image processing for translating a sensor captured image to a display image |
| CN105550997A (en) * | 2015-12-08 | 2016-05-04 | 天津津航计算技术研究所 | Three-dimensional matching image denoising method based on multiple transform domains |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110954921A (en) * | 2019-12-03 | 2020-04-03 | 浙江大学 | Laser radar echo signal-to-noise ratio improving method based on block matching 3D collaborative filtering |
| CN114509749A (en) * | 2022-04-19 | 2022-05-17 | 亿慧云智能科技（深圳）股份有限公司 | Indoor positioning detection system and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108876724B (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102184526B (en) | Natural image denoising method based on dictionary learning and block matching | |
| Hu et al. | Depth map denoising using graph-based transform and group sparsity | |
| CN104159003B (en) | A kind of cooperateed with based on 3D filters the video denoising method rebuild with low-rank matrix and system | |
| CN104504652A (en) | Image denoising method capable of quickly and effectively retaining edge and directional characteristics | |
| CN105574829A (en) | Adaptive bilateral filtering algorithm for polarized SAR image | |
| CN102147915A (en) | Method for restoring weighting sparse edge regularization image | |
| CN102073999A (en) | Natural image noise removal method based on dual redundant dictionary learning | |
| CN112581378A (en) | Image blind deblurring method and device based on significance intensity and gradient prior | |
| CN108876724A (en) | SAR images filter method based on self-adaptation three-dimensional Block- matching | |
| CN102722879A (en) | SAR (synthetic aperture radar) image despeckle method based on target extraction and three-dimensional block matching denoising | |
| CN110473153B (en) | Image blind restoration method based on fuzzy kernel estimation iterative structure preservation | |
| CN107301631B (en) | SAR image speckle reduction method based on non-convex weighted sparse constraint | |
| CN102222327A (en) | Image denoising method based on Treelet transformation and minimum mean-square error estimation | |
| Qiang | Image denoising based on Haar wavelet transform | |
| Luo et al. | Shadow removal based on clustering correction of illumination field for urban aerial remote sensing images | |
| Chen et al. | Depth map inpainting via sparse distortion model | |
| CN103839237B (en) | SAR image despeckling method based on SVD dictionary and linear minimum mean square error estimation | |
| Li et al. | Video denoising using shape-adaptive sparse representation over similar spatio-temporal patches | |
| Guo et al. | Image interpolation based on nonlocal self-similarity | |
| CN113793280A (en) | Real image noise reduction method combining local noise variance estimation and BM3D block matching | |
| Bai et al. | Medical image denoising based on improving K-SVD and block-matching 3D filtering | |
| CN108090878B (en) | Interferometric phase filtering method based on disparity map and compensation filter | |
| Ahn et al. | Haze removal using visible and infrared image fusion | |
| Kumar et al. | Segmentation of image with DT-CWT and NLM filtering using fuzzy c-means clustering | |
| Hu et al. | Hyperspectral image restoration using nonconvex hybrid regularization |
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 |