CN107301631A - A kind of SAR image method for reducing speckle that sparse constraint is weighted based on non-convex - Google Patents
A kind of SAR image method for reducing speckle that sparse constraint is weighted based on non-convex Download PDFInfo
- Publication number
- CN107301631A CN107301631A CN201710506574.XA CN201710506574A CN107301631A CN 107301631 A CN107301631 A CN 107301631A CN 201710506574 A CN201710506574 A CN 201710506574A CN 107301631 A CN107301631 A CN 107301631A
- Authority
- CN
- China
- Prior art keywords
- msub
- mrow
- munder
- image
- msubsup
- 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 65
- 239000011159 matrix material Substances 0.000 claims abstract description 32
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 7
- 239000000654 additive Substances 0.000 claims description 4
- 230000000996 additive effect Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000000452 restraining effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 2
- 230000001427 coherent effect Effects 0.000 abstract description 7
- 238000012545 processing Methods 0.000 abstract description 6
- 230000008602 contraction Effects 0.000 abstract description 3
- 230000000717 retained effect Effects 0.000 abstract description 3
- 238000002474 experimental method Methods 0.000 description 9
- 238000003384 imaging method Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 238000001914 filtration Methods 0.000 description 4
- 230000009467 reduction Effects 0.000 description 4
- 238000005457 optimization Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 230000001629 suppression Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- G06T5/70—
-
- 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
Abstract
The invention discloses a kind of SAR image method for reducing speckle that sparse constraint is weighted based on non-convex, belong to digital image processing techniques field.It utilizes similar image set of blocks in the transform domain as illustrated openness, is found by similar image set of blocks and singular value decomposition is carried out by similarity-rough set for each target image block first and obtains coefficient matrix, then non-convex Weighted Constraint is carried out to coefficient matrix, and coefficient matrix is estimated by threshold value contraction, make the coefficient matrix estimated closer to true coefficient, finally reconstruct drop spot result using the coefficient matrix of estimation;The present invention makes the image after drop spot effectively suppress coherent speckle noise while details is retained by the non-convex Weighted Constraint to coefficient matrix, has obtained more accurate drop spot image, it is easier to target identification, therefore drop spot available for SAR image.
Description
Technical field
The invention belongs to digital image processing techniques field, its non-convex weighting more particularly to based on image block set is sparse
The mage retrieval model method of constraint, for SAR image drop spot processing.
Background technology
Synthetic aperture radar (SAR) imaging is with its round-the-clock round-the-clock, and confrontation weather condition interference performance is strong, distance to
The high-resolution imaging characteristicses of orientation and to be widely used in mapping, forecast of natural calamity and battle reconnaissance etc. civilian with military side
Face, but because the distinctive imaging processes of SAR make the presence of serious coherent speckle noise in SAR image, easily cause Small object identification
Difficulty, therefore, it is necessary to carry out Speckle reduction to it before the processing such as subsequent singulation identification is carried out to SAR image.
Multiple-look technique, i.e. several subgraphs to Same Scene are mainly to the method for Speckle reduction before SAR image imaging
It is averaging processing, the suppression that this method can be preliminary to the progress of SAR image coherent spot, and most SAR image coherent spots
Suppressing method is concentrated mainly on after imaging, is generally divided into spatial domain and the major class of transform domain two.Spatial domain speckle suppression method is main
The distribution of SAR image environmental model and noise model distribution are analyzed, and binding signal estimation theory is filtered in spatial domain to image
Processing, wherein more classical method has Lee filtering, Frost filtering and Kuan filtering etc., but its Speckle reduction is limited in one's ability
And it is not enough to the holding capacity of image edge detailss.Transform domain filtering method starts from the development of wavelet technique and is introduced in SAR
In mage retrieval model, there are a series of multi-scale transform methods to be suggested again on this basis.In recent years, with the hair of sparse theory
Exhibition, the method that the openness and non local similitude based on image is reconstructed is increasingly becoming the focus of research.Because image exists
Transform domain has openness, and different zones have similar structure in image, can further press down in combination with both characteristics
Coherent speckle noise processed, in this kind of method, more classical for SAR-BM3D methods, it drops spot result at present still in compared with Gao Shui
It is flat, but this method easily produces artifact phenomenon in smooth area, and interference is easily brought in target identification.
The content of the invention
It is an object of the invention to drop the deficiency retained in spot image detail for existing SAR image, a kind of base is proposed
The SAR image method for reducing speckle of sparse constraint is weighted in non-convex.This method takes into full account the non local similitude and low-rank of SAR image
It is structural, non-convex Weighted Constraint is carried out to the coefficient matrix of similar image set of blocks, therefore, the method can make the SAR estimated
Image suppresses coherent speckle noise while a large amount of details inside image are retained.Comprise the following steps:
Step 1: non-convex weights the foundation of sparse constraint model
Logarithmic transformation is carried out to SAR image first, Multiplicative noise model is converted into Additive noise model, then to input
I-th of target image block x in imagei, similarity-rough set is carried out with all image blocks in its hunting zone, similarity is chosen most
Several high image blocks collectively form similar image set of blocks R with target image blockiX, wherein RiMatrix is extracted for image block,
Finally setting up non-convex weighting sparse constraint model is:
Wherein x and y represent true picture and initial pictures to be estimated, X respectivelyiFor the similar of true picture to be estimated
Image block set,Represent XiThe weighting p norms (0 < p < 1) of corresponding coefficient matrix, ω is weight vectors, and λ and η are
Balance every parameter.
Step 2: the decomposition and conversion of model
Restricted model in step one is decomposed, the subproblem on solving similar image set of blocks is converted into:
With the subproblem of Image Reconstruction:
The subproblem of similar image set of blocks, first the similar image set of blocks R to input are solved for formula (2)iX is carried out
Singular value decomposition, obtains corresponding coefficient matrix Γi, then will be weighted for the non-convex of similar image set of blocks in true picture
Sparse constraint model conversation is the non-convex Weighted Constraint model for similar image set of blocks coefficient matrix in true picture:
Wherein ΔiFor similar image set of blocks X in true picture to be estimatediCorresponding coefficient matrix, δjFor ΔiMiddle jth
Individual coefficient, ωjFor corresponding weight parameter.
Step 3: the estimation and the reconstruct of image of coefficient matrix
In the coefficient matrix Δ to step 2 Chinese style (4)iWhen being estimated, due to ΔiIn each coefficient it is relatively independent,
Therefore it is to the estimation model of each coefficient:
Wherein γjRepresent coefficient matrix ΓiIn j-th of coefficient, then using threshold value shrink each coefficient is estimated
Meter, after the estimate of each coefficient is obtained, you can the similar image set of blocks X in the true picture estimatedi, then
The subproblem of formula (3) Image Reconstruction is solved using formula (6):
And loop iteration is solved on similar image set of blocks XiWith estimation image x subproblem, until restraining or reaching
Iterations, then carries out exponential transform, the drop spot SAR image finally estimated to estimation image x.
The innovative point of the present invention is the low-rank characteristic that similar image set of blocks is utilized during spot drops in SAR image, to it
Coefficient matrix carries out non-convex Weighted Constraint;And coefficient matrix is estimated using threshold value contraction, make estimated result closer
Actual value, and this method is used for SAR image drop spot.
Beneficial effects of the present invention:With reference to image block, locally openness and non local similitude carries out similar image Block- matching
And singular value decomposition, improve rarefaction representation performance;Using non-convex Weighted Constraint coefficient matrix, make coefficient closer to actual value;
And every one dimensional system number is estimated using threshold value contraction, make the result of estimation more accurate, therefore the image finally estimated is not
Substantial amounts of details is only remained, the generation of artifact is also effectively inhibited, makes whole structure closer to true picture.
The main method for using emulation experiment of the invention is verified that all steps, conclusion are verified all on MATLAB9.0
Correctly.
Brief description of the drawings
Fig. 1 is the workflow block diagram of the present invention;
Fig. 2 is the spot SAR image to be dropped used during the present invention is emulated;
Fig. 3 is drop spot result figure of the PPB methods to Fig. 2;
Fig. 4 is drop spot result figure of the SAR-BM3D methods to Fig. 2;
Fig. 5 is drop spot result figure of the inventive method to Fig. 2.
Embodiment
Reference picture 1, the present invention is the SAR image method for reducing speckle that sparse constraint is weighted based on non-convex, and specific steps are included such as
Under:
Step 1: non-convex weights the foundation of sparse constraint model
SAR image is subjected to logarithmic transformation, its Multiplicative noise model is converted to Additive noise model:
On the basis of additive model, for each target image block x in imagei, with owning in its hunting zone
Image block carries out similarity-rough set, to meet relatively adopting for similarity between the multiplying property aspect of model of SAR image, two image blocks
With formula (8):
Wherein xi(k) image block x is representediInterior k-th of pixel value, chooses and its S-1 image block of similarity highest and mesh
Logo image block constitutes similar image set of blocks RiX, and set up non-convex weighting sparse constraint model by formula (1).
Step 2: the decomposition and conversion of model
After non-convex Weighted Constraint model is set up, model is decomposed into two subproblems according to formula (2) and formula (3), wherein
For the similar image set of blocks of the true picture in the formula of trying to achieve (2)Need to be to the similar image set of blocks R of input pictureiX according to
Carry out singular value decomposition:
SVD(RiX)=Ui·Γi·ViFormula (9)
Wherein ΓiFor RiThe corresponding coefficient matrixes of x, UiAnd ViRespectively left and right orthogonal transform matrix, then by formula (2)
Non-convex weighting sparse constraint model conversation for similar image set of blocks in true picture is to be directed in true picture in formula (4)
The non-convex Weighted Constraint model of similar image set of blocks coefficient matrix, wherein weight parameter ωjIt can be calculated and obtained by formula (7):
Wherein γjFor ΓiIn j-th of coefficient, c is to regard the different constants changed of number according to SAR image, and ε is avoids counting
It is worth a minimum positive number of overflow problem, formula (4) is further converted into scalar form:
It is converted into the optimization problem of the corresponding function sum of each coefficient.
Step 3: the estimation and the reconstruct of image of coefficient matrix
Because each coefficient is relatively independent in the optimization problem of step 2 Chinese style (11), therefore asking for formula (5) can be converted into
The optimization problem of each coefficient is solved, formula (5) is solved:
Wherein τ is threshold value, and δ is its iterative solution, and its threshold tau can take Derivative Characteristics during extreme value to solve according to formula (5) and obtain:
Iterative solution δ can be obtained after being restrained by formula (14) successive ignition:
δ(l+1)=| γj|-ωjp(δ(l))p-1Formula (14)
Wherein l is iterations, after each coefficient is gone out by this threshold value shrinkage estimation, you can obtain true picture
The estimate of similar image set of blocks coefficient matrix, then solves reconstructed image using formula (6), and loop iteration solve on
Similar image set of blocks XiWith estimation image x subproblem, until restraining or reaching iterations, then estimation image x is entered
Row index is converted, the drop spot SAR image finally estimated.
The effect of the present invention can be further illustrated by following emulation experiment:
First, experiment condition and content
Experiment condition:It is Fig. 2 to test the input picture used, and pixel size is 256 × 256.Each method for reducing speckle in experiment
All realized using MATLAB Programming with Pascal Language.
Experiment content:Under these experimental conditions, using PPB methods and SAR-BM3D methods and the inventive method progress pair
Than.Objective evaluation index homogeneity area variance and equivalent number ENL and the edge retention coefficient of whole image of spot ability drop
EPI integrates measurement.
Experiment 1:Drop spot is carried out to Fig. 2 respectively with the inventive method and existing PPB methods and SAR-BM3D methods to handle.
Wherein PPB methods are current SAR noise reductions more one of methods of classics, especially in homogeneity area, and it is Fig. 3 that it, which drops spot result,;
SAR-BM3D methods carry out estimation coefficient in transform domain using linear minimum mean-squared error, and famous with details reserve capability, and it drops
Spot result is Fig. 4.The inventive method sets tile size in experimentThe image block number S that similar image set of blocks is included
It is set to:S=80, final reconstruction result is Fig. 5.
Contrast PPB methods can be seen that PPB methods with the inventive method and be connect in the performance of smooth area with the inventive method
Closely, slightly it is better than this method, but the part details quilt in details relatively enriches the drop spot result in region in the smoothness of some regions
Transitions smooth, result is not so good as the inventive method;The result of SAR-BM3D methods is with the inventive method in details reserve capability
It is upper close, but there are a large amount of artifacts in smooth area, smooth effect is not as the inventive method and PPB methods;The inventive method is utilized
Non-convex weights sparse constraint method to enter coefficient matrix row constraint, and the estimation for realizing coefficient matrix is shunk using threshold value, makes
Drop spot result can not only retain most of details in original image, and the smooth effect of smooth area is also preferable, whole image
Visual effect it is good, be easy to the subsequent treatment of the SAR images such as object recognition.
The Indexes Comparison of the different method for reducing speckle of table 1
Table 1 gives corresponding variance when carrying out drop spot to two regions of Fig. 2 using distinct methods, ENL values and entirely schemed
As the situation of EPI values, wherein variance is smaller or the higher expression smooth area drop spot effect of ENL values is better, the more high certain journey of EPI values
Represent that edge details are kept as on degree better, therefore the drop spot result of SAR image should integrate the result of the two indexs.It can see
Go out the inventive method contrast other method to compare, than more prominent in terms of smooth and details holding, while details is kept
Coherent spot is inhibited, and PPB methods are showed preferably only in ENL values and variance, not as SAR-BM3D methods and this in EPI values
Inventive method, SAR-BM3D methods in EPI values then with PPB methods on the contrary, show preferable, and not as PPB in variance and ENL values
Method and this method, this is consistent with intuitively visual results.
Above-mentioned experiment shows that method for reducing speckle of the present invention restrained effectively relevant while a large amount of detailed information are remained
Spot noise, while visual effect and objective evaluation index are all preferable, it can be seen that it is effective that the present invention drops spot to SAR image.
Claims (1)
1. a kind of SAR image method for reducing speckle that sparse constraint is weighted based on non-convex, it is characterised in that comprise the following steps that:
Step 1: non-convex weights the foundation of sparse constraint model
Logarithmic transformation is carried out to SAR image first, Multiplicative noise model Additive noise model is converted into, then to input picture
In i-th of target image block xi, similarity-rough set is carried out with all image blocks in its hunting zone, similarity highest is chosen
Several image blocks collectively form similar image set of blocks R with target image blockiX, wherein RiMatrix is extracted for image block, finally
Setting up non-convex weighting sparse constraint model is:
<mrow>
<mo>(</mo>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mo>,</mo>
<msub>
<mover>
<mi>X</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
</msub>
<mo>)</mo>
<mo>=</mo>
<munder>
<mrow>
<mi>arg</mi>
<mi>min</mi>
</mrow>
<mrow>
<mi>x</mi>
<mo>,</mo>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
</mrow>
</munder>
<mo>|</mo>
<mo>|</mo>
<mi>x</mi>
<mo>-</mo>
<mi>y</mi>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<mi>&eta;</mi>
<munder>
<mo>&Sigma;</mo>
<mi>i</mi>
</munder>
<mo>(</mo>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<mi>&lambda;</mi>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mrow>
<msub>
<mi>S</mi>
<mi>p</mi>
</msub>
<mo>,</mo>
<mi>&omega;</mi>
</mrow>
<mi>p</mi>
</msubsup>
<mo>)</mo>
</mrow>
Wherein x and y represent true picture and initial pictures to be estimated, X respectivelyiFor the similar image of true picture to be estimated
Set of blocks,Represent XiThe weighting p norms (0 < p < 1) of corresponding coefficient matrix, ω is weight vectors, and λ and η are balance
Every parameter;
Step 2: the decomposition and conversion of model
Restricted model in step one is decomposed, the subproblem on solving similar image set of blocks is converted into:
<mrow>
<msub>
<mover>
<mi>X</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
</msub>
<mo>=</mo>
<munder>
<mrow>
<mi>arg</mi>
<mi>min</mi>
</mrow>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
</munder>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<mi>&lambda;</mi>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mrow>
<msub>
<mi>S</mi>
<mi>p</mi>
</msub>
<mo>,</mo>
<mi>&omega;</mi>
</mrow>
<mi>p</mi>
</msubsup>
</mrow>
With the subproblem of Image Reconstruction:
<mrow>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mo>=</mo>
<munder>
<mi>argmin</mi>
<mi>x</mi>
</munder>
<mo>|</mo>
<mo>|</mo>
<mi>x</mi>
<mo>-</mo>
<mi>y</mi>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<mi>&eta;</mi>
<munder>
<mo>&Sigma;</mo>
<mi>i</mi>
</munder>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
</mrow>
Subproblem for solving similar image set of blocks, first the similar image set of blocks R to inputiX carries out singular value point
Solution, obtains corresponding coefficient matrix Γi, then will weight sparse constraint for the non-convex of similar image set of blocks in true picture
Model conversation is the non-convex Weighted Constraint model for similar image set of blocks coefficient matrix in true picture:
<mrow>
<msub>
<mover>
<mi>&Delta;</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
</msub>
<mo>=</mo>
<munder>
<mrow>
<mi>arg</mi>
<mi>min</mi>
</mrow>
<msub>
<mi>&Delta;</mi>
<mi>i</mi>
</msub>
</munder>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>&Gamma;</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&Delta;</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<munder>
<mo>&Sigma;</mo>
<mi>j</mi>
</munder>
<msub>
<mi>&omega;</mi>
<mi>j</mi>
</msub>
<msup>
<msub>
<mi>&delta;</mi>
<mi>j</mi>
</msub>
<mi>p</mi>
</msup>
</mrow>
Wherein ΔiFor similar image set of blocks X in true picture to be estimatediCorresponding coefficient matrix, δjFor ΔiIn be for j-th
Number, ωjFor corresponding weight parameter;
Step 3: the estimation and the reconstruct of image of coefficient matrix
Coefficient matrix Δ in step 2iWhen being estimated, due to ΔiIn each coefficient it is relatively independent, therefore to each
The estimation model of coefficient is:
<mrow>
<msub>
<mover>
<mi>&delta;</mi>
<mo>^</mo>
</mover>
<mi>j</mi>
</msub>
<mo>=</mo>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<msub>
<mi>&delta;</mi>
<mi>j</mi>
</msub>
</munder>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>&delta;</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&gamma;</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msub>
<mi>&omega;</mi>
<mi>j</mi>
</msub>
<msup>
<msub>
<mi>&delta;</mi>
<mi>j</mi>
</msub>
<mi>p</mi>
</msup>
</mrow>
Wherein γjRepresent coefficient matrix ΓiIn j-th of coefficient, then using threshold value shrink each coefficient is estimated,
After the estimate for obtaining each coefficient, you can the similar image set of blocks X in the true picture estimatedi, then solve
The subproblem of Image Reconstruction:
<mrow>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mo>=</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>I</mi>
<mo>+</mo>
<mi>&eta;</mi>
<munder>
<mo>&Sigma;</mo>
<mi>i</mi>
</munder>
<msup>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mi>T</mi>
</msup>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>+</mo>
<mi>&eta;</mi>
<munder>
<mo>&Sigma;</mo>
<mi>i</mi>
</munder>
<msup>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mi>T</mi>
</msup>
<msub>
<mi>X</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
And loop iteration is solved on similar image set of blocks XiWith estimation image x subproblem, until restraining or reaching iteration time
Number, then carries out exponential transform, the drop spot SAR image finally estimated to estimation image x.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710506574.XA CN107301631B (en) | 2017-06-28 | 2017-06-28 | SAR image speckle reduction method based on non-convex weighted sparse constraint |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710506574.XA CN107301631B (en) | 2017-06-28 | 2017-06-28 | SAR image speckle reduction method based on non-convex weighted sparse constraint |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107301631A true CN107301631A (en) | 2017-10-27 |
CN107301631B CN107301631B (en) | 2020-09-18 |
Family
ID=60135069
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710506574.XA Active CN107301631B (en) | 2017-06-28 | 2017-06-28 | SAR image speckle reduction method based on non-convex weighted sparse constraint |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107301631B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109658340A (en) * | 2018-10-17 | 2019-04-19 | 南京航空航天大学 | The SAR image rapid denoising method saved based on RSVD and histogram |
CN112099010A (en) * | 2020-09-16 | 2020-12-18 | 中国人民解放军国防科技大学 | ISAR (inverse synthetic aperture radar) imaging method for target with micro-motion component based on structured non-convex low-rank representation |
CN112488960A (en) * | 2020-12-16 | 2021-03-12 | 中国人民解放军国防科技大学 | SAR image speckle suppression method based on boundary and context constraint |
CN112927165A (en) * | 2021-03-22 | 2021-06-08 | 重庆邮电大学 | SAR image speckle suppression method based on NSST domain three-dimensional block matching |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130191425A1 (en) * | 2012-01-20 | 2013-07-25 | Fatih Porikli | Method for Recovering Low-Rank Matrices and Subspaces from Data in High-Dimensional Matrices |
CN104933683A (en) * | 2015-06-09 | 2015-09-23 | 南昌大学 | Non-convex low-rank reconstruction method for rapid magnetic resonance (MR) imaging |
CN104952051A (en) * | 2015-07-07 | 2015-09-30 | 西安电子科技大学 | Low-rank image restoration method based on Gaussian mixture model |
CN105427264A (en) * | 2015-12-23 | 2016-03-23 | 重庆大学 | Image reconstruction method based on group sparsity coefficient estimation |
US20160156652A1 (en) * | 2012-04-20 | 2016-06-02 | Numerica Corporaition | Pattern detection in sensor networks |
-
2017
- 2017-06-28 CN CN201710506574.XA patent/CN107301631B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130191425A1 (en) * | 2012-01-20 | 2013-07-25 | Fatih Porikli | Method for Recovering Low-Rank Matrices and Subspaces from Data in High-Dimensional Matrices |
US20160156652A1 (en) * | 2012-04-20 | 2016-06-02 | Numerica Corporaition | Pattern detection in sensor networks |
CN104933683A (en) * | 2015-06-09 | 2015-09-23 | 南昌大学 | Non-convex low-rank reconstruction method for rapid magnetic resonance (MR) imaging |
CN104952051A (en) * | 2015-07-07 | 2015-09-30 | 西安电子科技大学 | Low-rank image restoration method based on Gaussian mixture model |
CN105427264A (en) * | 2015-12-23 | 2016-03-23 | 重庆大学 | Image reconstruction method based on group sparsity coefficient estimation |
Non-Patent Citations (3)
Title |
---|
JINGBO WEI ET AL.: "Nonlocal Low-Rank-Based Compressed Sensing for Remote Sensing Image Reconstruction", 《 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS 》 * |
林乐平: "基于过完备字典的非凸压缩感知理论与方法研究基于过完备字典的非凸压缩感知理论与方法研究", 《中国博士学位论文全文数据库信息科技辑》 * |
赵杰等: "基于非局部相似和低秩矩阵逼近的SAR图像去噪", 《计算机科学》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109658340A (en) * | 2018-10-17 | 2019-04-19 | 南京航空航天大学 | The SAR image rapid denoising method saved based on RSVD and histogram |
CN112099010A (en) * | 2020-09-16 | 2020-12-18 | 中国人民解放军国防科技大学 | ISAR (inverse synthetic aperture radar) imaging method for target with micro-motion component based on structured non-convex low-rank representation |
CN112099010B (en) * | 2020-09-16 | 2021-05-25 | 中国人民解放军国防科技大学 | ISAR (inverse synthetic aperture radar) imaging method for target with micro-motion component based on structured non-convex low-rank representation |
CN112488960A (en) * | 2020-12-16 | 2021-03-12 | 中国人民解放军国防科技大学 | SAR image speckle suppression method based on boundary and context constraint |
CN112927165A (en) * | 2021-03-22 | 2021-06-08 | 重庆邮电大学 | SAR image speckle suppression method based on NSST domain three-dimensional block matching |
Also Published As
Publication number | Publication date |
---|---|
CN107301631B (en) | 2020-09-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11238602B2 (en) | Method for estimating high-quality depth maps based on depth prediction and enhancement subnetworks | |
CN110119780B (en) | Hyper-spectral image super-resolution reconstruction method based on generation countermeasure network | |
CN106204468B (en) | A kind of image de-noising method based on ReLU convolutional neural networks | |
Cai et al. | Degradation-aware unfolding half-shuffle transformer for spectral compressive imaging | |
CN107301631A (en) | A kind of SAR image method for reducing speckle that sparse constraint is weighted based on non-convex | |
CN101950365B (en) | Multi-task super-resolution image reconstruction method based on KSVD dictionary learning | |
CN112233026A (en) | SAR image denoising method based on multi-scale residual attention network | |
CN101882304B (en) | Self-adaptive de-noising and characteristic enhancing method of SAR (Synthetic Aperture Radar) image | |
CN111667399A (en) | Method for training style migration model, method and device for video style migration | |
CN112001868A (en) | Infrared and visible light image fusion method and system based on generation of antagonistic network | |
CN107274360B (en) | A kind of high spectrum image denoising method based on Fisher dictionary learning, low-rank representation | |
CN103632153B (en) | Region-based image saliency map extracting method | |
CN107085835B (en) | Color image filtering method based on quaternary number Weighted Kernel Norm minimum | |
CN105427264A (en) | Image reconstruction method based on group sparsity coefficient estimation | |
CN104657951A (en) | Multiplicative noise removal method for image | |
CN102567973A (en) | Image denoising method based on improved shape self-adaptive window | |
CN116485934A (en) | Infrared image colorization method based on CNN and ViT | |
CN109064402A (en) | Based on the single image super resolution ratio reconstruction method for enhancing non local total variation model priori | |
CN107085826B (en) | Image super-resolution reconstruction method based on weighted overlap non-local regression prior | |
CN106204502A (en) | Based on mixing rank L0regularization fuzzy core method of estimation | |
CN103632372A (en) | Video saliency image extraction method | |
CN104978716A (en) | SAR image noise reduction method based on linear minimum mean square error estimation | |
CN106934398A (en) | Image de-noising method based on super-pixel cluster and rarefaction representation | |
CN103839237B (en) | SAR image despeckling method based on SVD dictionary and linear minimum mean square error estimation | |
CN116740485A (en) | Training method and device for lesion detection model, electronic equipment and storage medium |
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 |