CN105374017B - A kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution - Google Patents
A kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution Download PDFInfo
- Publication number
- CN105374017B CN105374017B CN201510862959.0A CN201510862959A CN105374017B CN 105374017 B CN105374017 B CN 105374017B CN 201510862959 A CN201510862959 A CN 201510862959A CN 105374017 B CN105374017 B CN 105374017B
- Authority
- CN
- China
- Prior art keywords
- mrow
- msubsup
- mtd
- pixel
- mtr
- 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.)
- Expired - Fee Related
Links
Abstract
The invention discloses a kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution, by carrying out Polarization target decomposition to the polarimetric SAR image data of input, obtains scattering vector;Non local method is utilized to the polarization SAR data of input, obtains the weights of other pixels in each pixel and its search window;According to the distribution character of the polarization decomposing of polarization SAR data vector, try to achieve the measure formulas and threshold value of Polarization Characteristics Similarity, similarity measurement is carried out for the pixel in the search window of each pixel, the similar set of each pixel is found, weights formula is modified;Each pixel is filtered using final weights formula, obtains filtered polarimetric SAR image data;The present invention solves the problems, such as that filtering method can not keep image detail information and scattering properties very well so that can also keep the details and scattering properties of image well while Speckle reduction is carried out to Polarimetric SAR Image.
Description
Technical field
The invention belongs to technical field of image processing, can be used for the Speckle reduction of Polarimetric SAR Image.It is specifically a kind of
The Polarimetric SAR Image filtering method being distributed with reference to polarization decomposing vector statistical.
Background technology
Polarimetric synthetic aperture radar (Pol-SAR) is to develop to come on the basis of SAR, and the target information that it is obtained is more
It is abundant, it is obtained for and is widely applied in military and civil field.But coherent speckle noise in Polarimetric SAR Image being present, this gives
Follow-up image interpretation and analysis brings difficulty, so, seek to that Polarimetric SAR Image is carried out to drop spot processing first.
At present, there are many polarization SAR method for reducing speckle, representative has polarimetric whitening filter (PWF), polarization exquisiteness
The methods of Lee is filtered, although these methods have certain drop spot effect, at edge, the field such as lines to Polarimetric SAR Image
It lost many details so that image thickens unclear.There are recent years many scholars that non local thought is incorporated into pole
Change in SAR drop spots, it is proposed that many methods, as the non-local mean of Yang Jian et al. polarization SARs proposed filters --- ---
Pretest methods, this method are to calculate weights using the similar block of image, and this method make use of the structure of image to believe well
Breath, also there is certain drop spot effect to image, but many small details of the missing image after drop spot, such as target point, carefully
Also there is blooming in lines etc., edge.
The content of the invention
The purpose of the present invention is to be directed to problem above, proposes a kind of polarization SAR of combination polarization decomposing vector statistical distribution
Image filtering method, mainly solves the detailed information that existing filtering algorithm can not keep image well when to image filtering
And the problem of scattering properties.
The technical scheme is that a kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution,
Comprise the following steps:
Step 1:Input the coherence matrix T of polarization SAR data;
Step 2:Coherence matrix T is carried out to mix the decomposition of four components, i.e. HPD is decomposed, and each pixel is divided into four kinds and dissipated
Type is penetrated, i.e.,:Surface scattering ps, even scattering pd, volume scattering pv and spiral scattering ph, the scattering properties of each pixel are used
One 1 × 3 vector representation, then pixel i scattering properties be:veci=[psi,pdi,pvi+phi];
Step 3:Using non local technology, the weight w of other pixels j in each pixel i and its neighborhood is tried to achieve
(i,j);
Step 4:According to the distribution character of the polarization decomposing of polarization SAR data vector, the measurement for trying to achieve Polarization Characteristics Similarity is public
Formula and threshold value, and then each pixel i similitude pixels in its neighborhood window are found, weights are modified, obtained most
Whole weights formula;
Step 5:Each pixel is estimated using final weights formula, obtains whole Polarimetric SAR Image filter
Coherence matrix after ripple
Step 6:Using Pauli decomposition methods by filtered coherence matrixSynthesize pcolor.
Above-mentioned steps two comprise the following steps:
201:Coherence matrix T is subjected to directional process, obtains the coherence matrix T after orientation0,
202:According to coherence matrix T0HPD can be obtained and decompose vector v ec,
203:Pv and ph is solved,
WhenWhen:
Wherein, Re () and Im () represents real and imaginary part respectively, and r definition is:
WhenWhen:
204:Target end decomposition algorithm, each pixel i resolves into one 3 × 1 scattering vector, labeled as veci=
[psi,pdi,pvi+phi]。
Above-mentioned steps 201 comprise the following steps:
301:Go out the orientation angle θ of each pixel using Huynen parameter Estimations;
302:Utilize formula
Obtain the coherence matrix T after orientation angle0。
Above-mentioned steps three comprise the following steps:
401:Image block I represents to expand 7 × 7 image block centered on pixel i, by centered on pixel i
Each pixel j in 15 × 15 search window Ω1,j2,...,jNCentered on expand respective 7 × 7 image block, respectively
Labeled as image block J1,J2,...,JN, N is the pixel number in search window, is 225;
402:Calculate image block I and image block JnBetween similarity measurement matrix HI,Jn, wherein n=1,2 ..., N,
As pixel i and pixel jnBetween similarity measurements moment matrix,
Wherein, M is the number of pixel in image block, is that 49, L is that regarding for image counts, | Ik| it is k-th of image block I
The determinant of pixel matrix,For image block JnK-th of pixel matrix determinant;
403:According to similarity measurements moment matrix, weight function w (i, j are calculatedn),
Wherein, HtFor threshold value,K is an adjustable parameter, takes 20, L to regard number for image, M is image
The number of pixel, Z in blockiTo normalize weights, Zi=∑j∈s(i)W (i, j), s (i) represent search window where pixel i.
Above-mentioned steps four comprise the following steps:
501:Pixel i scattering properties vector, which marks, isi=[psi,pdi,pvi+phi], each picture in search window Ω
Vegetarian refreshments j1,j2,...,jNScattering vector mark forWhereinn
=1,2 ..., N, calculates pixel i and pixel jnThe absolute value of difference on three components, as similitude
Parameter:
502:The absolute value of three differences of two pixels that 501 are obtained is compared with threshold value respectively, ifThen pixel jnIt is right
In pixel i final weights W (i, jn)=w (i, jn);Otherwise, pixel jnIt is not just pixel i similitude, pixel
jnIt is not involved in pixel i estimation, i.e. W (i, jn)=0, obtain final weights formula;Wherein, on three components difference it is exhausted
The threshold value of value is counted to obtain by simulation polarization SAR data, the distribution that the difference on three components of similitude is obeyed is near
Like being normal distribution, the standard deviation of difference on three components is obtained by mass data, P (μ -2 δ are known by the property of normal distribution
The δ of < x≤μ+2)=0.954, threshold value is set to twice of standard deviation, the threshold value for obtaining absolute difference on each component is:TH1
=2* (0.992L-0.4994·psi+0.3146·L-1.1246),
TH2=2* (1.0010L-0.4997·pdi+0.0150L-0.7533),
TH3=2* (0.8639L-0.4321(pvi+phi)+0.0293L0.2038- 0.0242), L is that image regards number.
Above-mentioned steps five comprise the following steps:
601:Utilize the pixel j in pixel i search windows1,j2,...,jNIt is weighted averagely with weights, obtains pixel
Point i filter resultI.e.1≤n≤N, whereinIt is normalized parameter;
602:Each pixel is handled by step 601, obtains the filtered coherence matrix of entire image
Beneficial effects of the present invention:The present invention carries out goal decomposition to the polarization SAR data of input, obtains scattering vector;
Polarization SAR data are utilized with non local method, the weights between pixel in the pixel asked and its search window;According to polarization
The distribution character of the polarization decomposing vector of SAR data, tries to achieve the measure formulas and threshold value of Polarization Characteristics Similarity, and then try to achieve pixel
Similitude in its search window, weights formula is modified;Using final weights formula to data filtering process.With with
Preceding polarization SAR filtering technique is compared, and the present invention has advantages below:
1. the implementation process of the present invention is simple.
2. the present invention passes through number using the absolute value of difference on three components of pixel point scattering vector as Similarity Parameter
According to threshold value is fitted, the similar pixel of pixel point is selected by Similarity Parameter and threshold value, makes the similar pixel point of selection
It is more accurate.
3. the present invention is combined by the method for measuring similarity drawn using scattered partion vector and with non-local mean, right
Also the detailed information and scattering properties of image can be kept while Polarimetric SAR Image is filtered well.
The present invention is described in further details below with reference to accompanying drawing.
Brief description of the drawings
Fig. 1 is the flow chart of the present invention;
Fig. 2 is two groups of polarization SAR datagrams inputting of the present invention, first group of polarization SAR data wherein shown in Fig. 2 (a)
Ten for C-band regard the polarization SAR data in Canadian Ottawa areas;Second group of polarization SAR data shown in Fig. 2 (b) are L ripples
The four of section regard the polarization SAR data in San Francisco areas;
Fig. 3 is the result being filtered with existing exquisite the Lee filtering that polarizes, Pretest filtering and the present invention to Fig. 2 (a)
Figure;
Fig. 4 is the result being filtered with existing exquisite the Lee filtering that polarizes, Pretest filtering and the present invention to Fig. 2 (b)
Figure.
Embodiment
The invention provides a kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution, will polarize
Method for measuring similarity and non-local mean are combined, after the weights for the pixel asked using non local technology, according to pole
Change the distribution character of the polarization decomposing vector of SAR data, try to achieve similar pixel point of the pixel in its neighborhood, and weights are entered
Row modification, makes weights more accurate.
The present invention carries out goal decomposition to the polarization SAR data of input first, obtains scattering vector;To polarization SAR data
Using non local method, the weights between pixel in the pixel asked and its search window;According to the polarization of polarization SAR data
The distribution character of vector is decomposed, tries to achieve the measure formulas and threshold value of Polarization Characteristics Similarity, and then the pixel asked is in its search window
Similitude, weights formula is modified;Using final weights formula to data filtering process.
Referring to the drawings 1, the Polarimetric SAR Image filtering side of this combination polarization decomposing vector statistical distribution provided by the invention
Method, detailed step are:
Step 1:Input the coherence matrix T of polarization SAR data;
Step 2:Coherence matrix T is carried out to mix four components (HPD) decomposition, each pixel is divided into four kinds of scattering classes
Type, surface scattering (ps), even scattering (pd), volume scattering (pv) and spiral scattering (ph), each pixel i scattering properties
With 1 × 3 vector representation, i.e. veci=[psi,pdi,pvi+phi];Specifically include following steps:
(1) orientation angle is removed, each pixel i orientation angle θ is gone out first by Huynen parameter Estimationsi, utilize formula
Obtain the relevant square after pixel i orientations
Battle array Ti 0, and then obtain the coherence matrix T of view picture figure0,
(2) according to coherence matrix T0HPD can be obtained and decompose vector v ec,
(3) pv and ph is solved,
WhenWhen:
Wherein, Re () and Im () represents real and imaginary part respectively, and r definition is:
WhenWhen:
(4) target end decomposition algorithm, each pixel i resolves into one 3 × 1 scattering vector, labeled as veci=
[psi,pdi,pvi+phi]。
Step 3:Using non local technology, each pixel i asked and interior other pixels j of its neighborhood weight w
(i, j), specifically include following steps:
(1) image block I represents to expand 7 × 7 image block centered on pixel i, by 15 centered on pixel i
Each pixel j in × 15 search window Ω1,j2,...,jNCentered on expand respective 7 × 7 image block, mark respectively
It is designated as image block J1,J2,...,JN, N is the pixel number in search window, is 225;
(2) image block I and image block J are calculatednSimilarity measurement matrix H between (n=1,2 ..., N)I,Jn, it is
Pixel i and pixel jnBetween similarity measurements moment matrix,
Wherein, M is the number of pixel in image block, is that 49, L is that regarding for image counts, | Ik| it is k-th of image block I
The determinant of pixel matrix,For image block JnK-th of pixel matrix determinant;
(3) according to similarity measurements moment matrix, weight function w (i, j are calculatedn),
Wherein, HtFor threshold value,K is an adjustable parameter, takes 20, L to regard number for image, M is image
The number of pixel, Z in blockiTo normalize weights, Zi=∑j∈s(i)W (i, j), s (i) represent search window where pixel i.
Step 4:According to the distribution character of the polarization decomposing of polarization SAR data vector, the measurement for the Polarization Characteristics Similarity tried to achieve
Formula and threshold value, and then each pixel i similitude pixels in its neighborhood window are looked for obtaining, weights are modified;Specifically
Comprise the following steps:
(1) pixel i scattering properties vector mark isi=[psi,pdi,pvi+phi], each picture in search window Ω
Vegetarian refreshments j1,j2,...,jNScattering vector mark forWhereinn
=1,2 ..., N, calculates pixel i and pixel jnThe absolute value of difference on three components as Similarity Parameter,
(2) absolute value of three differences of two pixels for obtaining (1) is compared with threshold value respectively, such asThen pixel jnIt is right
In pixel i final weights W (i, jn)=w (i, jn);Otherwise, pixel jnIt is not just pixel i similitude, pixel
jnIt is not involved in pixel i estimation, i.e. W (i, jn)=0, obtain final weights formula;Wherein, on three components difference it is exhausted
The threshold value of value is counted to obtain by simulation polarization SAR data, the distribution that the difference on three components of similitude is obeyed is near
Like being normal distribution, the distribution of the standard deviation of difference on three components is obtained by mass data, by the Property P of normal distribution
(δ of μ -2 δ < x≤μ+2)=0.954, threshold value is set to twice of standard deviation, obtains the threshold value of absolute difference on each component
For:TH1=2* (0.992L-0.4994·psi+0.3146·L-1.1246),
TH2=2* (1.0010L-0.4997·pdi+0.0150L-0.7533),
TH3=2* (0.8639L-0.4321(pvi+phi)+0.0293L0.2038- 0.0242), L is that image regards number.
Step 5:Utilize the pixel j in pixel i search windows1,j2,...,jNIt is weighted averagely, obtains with weights
Pixel i filter resultI.e.1≤n≤N, whereinIt is normalization ginseng
Number.Each pixel is so done, you can obtains the filtered coherence matrix of view picture figure
Step 6:Using Pauli decomposition methods by filtered coherence matrixSynthesize pcolor.
Effect of the present invention can further be confirmed by following experiment:
One, experiment conditions and content
Experiment condition:It is core21.86GHZ, carried out in internal memory 2G, WINDOWS7 system using MATLAB2008 in CPU
Emulation.
Experiment content:The present invention is from the polarization SAR data and four that two groups of polarization SAR data are the ten Ottawa areas regarded
Depending on San Francisco area polarization SAR data.Control methods is exquisite polarization Lee filtering and Pretest filtering.
Two, experimental results
Fig. 3 is the result being filtered with exquisite polarization Lee filtering, Pretest filtering and the present invention to Fig. 2 (a).Wherein
Fig. 3 (a) is exquisite polarization Lee filter results, and Fig. 3 (b) Pretest filter results, Fig. 3 (c) is filter result of the present invention.From figure
3 (a) is visible, and exquisiteness polarization Lee filtering has certain drop spot effect, but obscuring occurs in marginal portion;From Fig. 3 (b),
Pretest filters the noise that can farthest suppress homogeneous region, but excessively smooth phenomenon also occurs;Can from Fig. 3 (c)
See, while the present invention can suppress noise and the detailed information such as point target, lines, edge can be kept, kept well
The scattering properties of image, it can specifically see the part marked in figure with oval and square frame.
Fig. 4 is the result being filtered with exquisite polarization Lee filtering, Pretest filtering and the present invention to Fig. 2 (b).Wherein
Fig. 4 (a) is exquisite polarization Lee filter results, and Fig. 4 (b) Pretest filter results, Fig. 4 (c) is filter result of the present invention.From figure
4 (a) is visible, and exquisiteness polarization Lee filtering has certain drop spot effect, but blooming occurs in wood land;From Fig. 4
(b) visible, Pretest filtering can farthest suppress the noise of homogeneous region, but lost many detailed information;From figure
4 (c) is visible, and the present invention can be good at keeping the detailed information such as point target, lines, edge, maintains dissipating for image well
Penetrate characteristic.
In summary, this polarization SAR non-local mean filter based on Polarization Characteristics Similarity measure that the present invention provides
Wave technology, compared to above-mentioned two methods, it can be kept well in image while being filtered to Polarimetric SAR Image
The detailed information such as point target, lines, edge, the scattered information of image is maintained well.
There is no the known conventional means of the part category industry described in detail in present embodiment, do not chat one by one here
State.It is exemplified as above be only to the present invention for example, do not form the limitation to protection scope of the present invention, it is every with this
Same or analogous design is invented to belong within protection scope of the present invention.
Claims (4)
1. a kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution, it is characterised in that including following step
Suddenly:
Step 1:Input the coherence matrix T of polarization SAR data;
Step 2:Coherence matrix T is carried out to mix the decomposition of four components, i.e. HPD is decomposed, and each pixel is divided into four kinds of scattering classes
Type, i.e.,:Surface scattering ps, even scattering pd, volume scattering pv and spiral scattering ph, the scattering properties of each pixel is with one
1 × 3 vector representation, then pixel i scattering properties be:veci=[psi,pdi,pvi+phi];
The step 2 comprises the following steps:
201:Coherence matrix T is subjected to directional process, obtains the coherence matrix T after orientation0,
<mrow>
<msup>
<mi>T</mi>
<mn>0</mn>
</msup>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msubsup>
<mi>T</mi>
<mn>11</mn>
<mn>0</mn>
</msubsup>
</mtd>
<mtd>
<msubsup>
<mi>T</mi>
<mn>12</mn>
<mn>0</mn>
</msubsup>
</mtd>
<mtd>
<msubsup>
<mi>T</mi>
<mn>13</mn>
<mn>0</mn>
</msubsup>
</mtd>
</mtr>
<mtr>
<mtd>
<msubsup>
<mi>T</mi>
<mn>21</mn>
<mn>0</mn>
</msubsup>
</mtd>
<mtd>
<msubsup>
<mi>T</mi>
<mn>22</mn>
<mn>0</mn>
</msubsup>
</mtd>
<mtd>
<msubsup>
<mi>T</mi>
<mn>23</mn>
<mn>0</mn>
</msubsup>
</mtd>
</mtr>
<mtr>
<mtd>
<msubsup>
<mi>T</mi>
<mn>31</mn>
<mn>0</mn>
</msubsup>
</mtd>
<mtd>
<msubsup>
<mi>T</mi>
<mn>32</mn>
<mn>0</mn>
</msubsup>
</mtd>
<mtd>
<msubsup>
<mi>T</mi>
<mn>33</mn>
<mn>0</mn>
</msubsup>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>;</mo>
</mrow>
202:According to coherence matrix T0HPD can be obtained and decompose vector v ec,
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<mi>v</mi>
<mi>e</mi>
<mi>c</mi>
<mo>=</mo>
<mo>&lsqb;</mo>
<mi>p</mi>
<mi>s</mi>
<mo>,</mo>
<mi>p</mi>
<mi>d</mi>
<mo>,</mo>
<mi>p</mi>
<mi>v</mi>
<mo>+</mo>
<mi>p</mi>
<mi>h</mi>
<mo>&rsqb;</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<mrow>
<mo>&lsqb;</mo>
<mrow>
<msubsup>
<mi>T</mi>
<mn>11</mn>
<mn>0</mn>
</msubsup>
<mo>,</mo>
<msubsup>
<mi>T</mi>
<mn>22</mn>
<mn>0</mn>
</msubsup>
<mo>,</mo>
<mi>p</mi>
<mi>v</mi>
<mo>+</mo>
<mi>p</mi>
<mi>h</mi>
</mrow>
<mo>&rsqb;</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>;</mo>
</mrow>
203:Pv and ph is solved,
WhenWhen:
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>p</mi>
<mi>v</mi>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mn>15</mn>
<mo>/</mo>
<mn>4</mn>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>T</mi>
<mn>33</mn>
<mn>0</mn>
</msubsup>
<mo>-</mo>
<mo>|</mo>
<mi>Im</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>T</mi>
<mn>23</mn>
<mn>0</mn>
</msubsup>
<mo>)</mo>
</mrow>
<mo>|</mo>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
<mtd>
<mrow>
<mo>|</mo>
<mi>r</mi>
<mo>|</mo>
<mo>></mo>
<mn>2</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>4</mn>
<mo>&CenterDot;</mo>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>T</mi>
<mn>33</mn>
<mn>0</mn>
</msubsup>
<mo>-</mo>
<mo>|</mo>
<mi>Im</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>T</mi>
<mn>23</mn>
<mn>0</mn>
</msubsup>
<mo>)</mo>
</mrow>
<mo>|</mo>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
<mtd>
<mrow>
<mo>|</mo>
<mi>r</mi>
<mo>|</mo>
<mo>&le;</mo>
<mn>2</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>p</mi>
<mi>h</mi>
<mo>=</mo>
<mn>2</mn>
<mo>&CenterDot;</mo>
<mo>|</mo>
<mi>Im</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>T</mi>
<mn>23</mn>
<mn>0</mn>
</msubsup>
<mo>)</mo>
</mrow>
<mo>|</mo>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Wherein, Re () and Im () represents real and imaginary part respectively, and r definition is:
<mrow>
<mi>r</mi>
<mo>=</mo>
<mn>10</mn>
<mo>&times;</mo>
<mi>l</mi>
<mi>o</mi>
<mi>g</mi>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<msubsup>
<mi>T</mi>
<mn>11</mn>
<mn>0</mn>
</msubsup>
<mo>+</mo>
<msubsup>
<mi>T</mi>
<mn>22</mn>
<mn>0</mn>
</msubsup>
<mo>-</mo>
<mn>2</mn>
<mo>&CenterDot;</mo>
<mi>Re</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>T</mi>
<mn>12</mn>
<mn>0</mn>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msubsup>
<mi>T</mi>
<mn>11</mn>
<mn>0</mn>
</msubsup>
<mo>+</mo>
<msubsup>
<mi>T</mi>
<mn>22</mn>
<mn>0</mn>
</msubsup>
<mo>+</mo>
<mn>2</mn>
<mo>&CenterDot;</mo>
<mi>Re</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>T</mi>
<mn>12</mn>
<mn>0</mn>
</msubsup>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
</mrow>
WhenWhen:
<mrow>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>p</mi>
<mi>v</mi>
<mo>=</mo>
<mn>4</mn>
<mo>&CenterDot;</mo>
<msubsup>
<mi>T</mi>
<mn>33</mn>
<mn>0</mn>
</msubsup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>p</mi>
<mi>h</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>;</mo>
</mrow>
204:Target end decomposition algorithm, each pixel i resolves into one 3 × 1 scattering vector, labeled as veci=[psi,
pdi,pvi+phi];
Step 3:Using non local technology, the weight w (i, j) of other pixels j in each pixel i and its neighborhood is tried to achieve;
Step 4:According to the distribution character of the polarization decomposings of polarization SAR data vector, try to achieve Polarization Characteristics Similarity measure formulas and
Threshold value, and then similar pixel points of each pixel i in its neighborhood window is found, weights are modified, obtained final
Weights formula;
Step 5:Each pixel is estimated using final weights formula, after obtaining whole Polarimetric SAR Image filtering
Coherence matrix
Step 6:Using Pauli decomposition methods by filtered coherence matrixSynthesize pcolor.
2. a kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution according to claim 1, its
It is characterised by, the step 201 comprises the following steps:
301:Go out the orientation angle θ of each pixel using Huynen parameter Estimations;
302:Utilize formula
Obtain the coherence matrix T after orientation angle0。
3. a kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution according to claim 1, its
It is characterised by, the step 3 comprises the following steps:
401:Image block I represent expanded centered on pixel i 7 × 7 image block, by 15 centered on pixel i ×
Each pixel j in 15 search window Ω1,j2,...,jNCentered on expand respective 7 × 7 image block, mark respectively
For image block J1,J2,...,JN, N is the pixel number in search window, is 225;
402:Calculate image block I and image block JnBetween similarity measurements moment matrixWherein n=1,2 ..., N, it is
Pixel i and pixel jnBetween similarity measurements moment matrix,
<mrow>
<msub>
<mi>H</mi>
<mrow>
<mi>I</mi>
<mo>,</mo>
<msup>
<mi>J</mi>
<mi>n</mi>
</msup>
</mrow>
</msub>
<mo>=</mo>
<mn>6</mn>
<mi>M</mi>
<mo>&CenterDot;</mo>
<mi>L</mi>
<mo>&CenterDot;</mo>
<mi>l</mi>
<mi>n</mi>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mi>L</mi>
<mo>&CenterDot;</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>M</mi>
</munderover>
<mrow>
<mo>(</mo>
<mi>ln</mi>
<mo>|</mo>
<msub>
<mi>I</mi>
<mi>k</mi>
</msub>
<mo>|</mo>
<mo>+</mo>
<mi>l</mi>
<mi>n</mi>
<mo>|</mo>
<msubsup>
<mi>J</mi>
<mi>k</mi>
<mi>n</mi>
</msubsup>
<mo>|</mo>
<mo>-</mo>
<mn>2</mn>
<mi>ln</mi>
<mo>|</mo>
<msub>
<mi>I</mi>
<mi>k</mi>
</msub>
<mo>+</mo>
<msubsup>
<mi>J</mi>
<mi>k</mi>
<mi>n</mi>
</msubsup>
<mo>|</mo>
<mo>)</mo>
</mrow>
</mrow>
Wherein, M is the number of pixel in image block, is that 49, L is that regarding for image counts, | Ik| it is image block I k-th of pixel
Determinant of a matrix value,For image block JnK-th of pixel matrix determinant;
403:According to similarity measurements moment matrix, weight function w (i, j are calculatedn),
<mrow>
<mi>w</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<msup>
<mi>j</mi>
<mi>n</mi>
</msup>
</mrow>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mfrac>
<mn>1</mn>
<msub>
<mi>Z</mi>
<mi>i</mi>
</msub>
</mfrac>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mrow>
<mo>-</mo>
<mfrac>
<mrow>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<msup>
<mi>j</mi>
<mi>n</mi>
</msup>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<msub>
<mi>H</mi>
<mi>t</mi>
</msub>
</mfrac>
</mrow>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<msup>
<mi>j</mi>
<mi>n</mi>
</msup>
</mrow>
<mo>)</mo>
</mrow>
<mo>></mo>
<msub>
<mi>H</mi>
<mi>t</mi>
</msub>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>i</mi>
<mo>,</mo>
<msup>
<mi>j</mi>
<mi>n</mi>
</msup>
</mrow>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<msub>
<mi>H</mi>
<mi>t</mi>
</msub>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, HtFor threshold value,K is an adjustable parameter, takes 20, L to regard number for image, and M is in image block
The number of pixel, ZiTo normalize weights, Zi=∑j∈s(i)W (i, j), s (i) represent search window where pixel i.
4. a kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution according to claim 1, its
It is characterised by, the step 4 comprises the following steps:
501:Pixel i scattering properties vector, which marks, isi=[psi,pdi,pvi+phi], each pixel in search window Ω
Point j1,j2,...,jNScattering vector mark forWhereinn
=1,2 ..., N, calculates pixel i and pixel jnThe absolute value of difference on three components, as similitude
Parameter:
502:The absolute value of three differences of two pixels that 501 are obtained is compared with threshold value respectively, ifThen pixel jnIt is right
In pixel i final weights W (i, jn)=w (i, jn);Otherwise, pixel jnIt is not pixel i similar pixel point, pixel
Point jnIt is not involved in pixel i estimation, i.e. W (i, jn)=0, obtain final weights formula;Wherein, difference on three components
The threshold value of absolute value is:
TH1=2* (0.992L-0.4994·psi+0.3146·L-1.1246),
TH2=2* (1.0010L-0.4997·pdi+0.0150L-0.7533),
TH3=2* (0.8639L-0.4321(pvi+phi)+0.0293L0.2038- 0.0242), L is that image regards number.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510862959.0A CN105374017B (en) | 2015-11-30 | 2015-11-30 | A kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510862959.0A CN105374017B (en) | 2015-11-30 | 2015-11-30 | A kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105374017A CN105374017A (en) | 2016-03-02 |
CN105374017B true CN105374017B (en) | 2018-03-30 |
Family
ID=55376189
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510862959.0A Expired - Fee Related CN105374017B (en) | 2015-11-30 | 2015-11-30 | A kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105374017B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106203522B (en) * | 2016-07-15 | 2019-03-26 | 西安电子科技大学 | Hyperspectral image classification method based on three-dimensional non-local mean filtering |
CN108961284A (en) * | 2018-06-12 | 2018-12-07 | 中国电子科技集团公司第二十九研究所 | SAR image building extracting method, equipment and the storage medium of side lobe effect pollution |
CN109087257B (en) * | 2018-07-25 | 2020-08-07 | 武汉科技大学 | Airspace increment image filtering method based on parameter estimation framework |
CN110363105B (en) * | 2019-06-25 | 2021-12-10 | 电子科技大学 | Method for inhibiting speckle of fully-polarized SAR image |
CN112419198B (en) * | 2020-11-27 | 2024-02-02 | 中国矿业大学 | Non-local mean weighting method for SAR interferogram filtering |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101344587A (en) * | 2008-08-15 | 2009-01-14 | 哈尔滨工业大学 | Multi-component decomposition method used for polarization synthetic aperture radar image |
CN104240200A (en) * | 2014-09-02 | 2014-12-24 | 西安电子科技大学 | Polarimetric SAR speckle suppression method based on combination of scattering model and non-local mean values |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2828687B1 (en) * | 2012-03-23 | 2020-08-05 | Windar Photonics A/S | Multiple directional lidar system |
-
2015
- 2015-11-30 CN CN201510862959.0A patent/CN105374017B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101344587A (en) * | 2008-08-15 | 2009-01-14 | 哈尔滨工业大学 | Multi-component decomposition method used for polarization synthetic aperture radar image |
CN104240200A (en) * | 2014-09-02 | 2014-12-24 | 西安电子科技大学 | Polarimetric SAR speckle suppression method based on combination of scattering model and non-local mean values |
Non-Patent Citations (1)
Title |
---|
极化SAR相干斑抑制的非局部加权最小均方误差滤波算法;马晓双等;《中国图象图形学报》;20150131;第20卷(第1期);140-150 * |
Also Published As
Publication number | Publication date |
---|---|
CN105374017A (en) | 2016-03-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105374017B (en) | A kind of Polarimetric SAR Image filtering method of combination polarization decomposing vector statistical distribution | |
US10635929B2 (en) | Saliency-based method for extracting road target from night vision infrared image | |
CN108550161B (en) | Scale self-adaptive kernel-dependent filtering rapid target tracking method | |
Cao et al. | An Unsupervised Segmentation With an Adaptive Number of Clusters Using the $ SPAN/H/\alpha/A $ Space and the Complex Wishart Clustering for Fully Polarimetric SAR Data Analysis | |
CN103279957B (en) | A kind of remote sensing images area-of-interest exacting method based on multi-scale feature fusion | |
CN105528589A (en) | Single image crowd counting algorithm based on multi-column convolutional neural network | |
DE112016006873T5 (en) | Capture people in images using depth information | |
CN102663405B (en) | Prominence and Gaussian mixture model-based method for extracting foreground of surveillance video | |
CN105718957A (en) | Polarized SAR image classification method based on nonsubsampled contourlet convolutional neural network | |
CN105335975B (en) | Polarization SAR image segmentation method based on low-rank decomposition and statistics with histogram | |
CN109948593A (en) | Based on the MCNN people counting method for combining global density feature | |
Borghys et al. | Hyperspectral anomaly detection: Comparative evaluation in scenes with diverse complexity | |
CN103020919A (en) | Polarimetric SAR (synthetic aperture radar) phase speckled noise suppression method based on non-local Lee | |
CN103903012A (en) | Polarimetric SAR data classifying method based on orientation object and support vector machine | |
CN105354824B (en) | DP-CFAR detection method based on extracted region | |
US9183671B2 (en) | Method for accelerating Monte Carlo renders | |
CN103093432B (en) | Polarized synthetic aperture radar (SAR) image speckle reduction method based on polarization decomposition and image block similarity | |
CN103020959A (en) | Gravity model-based oceanic front information extraction method | |
DE112016006921T5 (en) | Estimation of human orientation in images using depth information | |
CN103268498A (en) | Method for area-of-interest blurred image semantic comprehension | |
CN104240200B (en) | Based on the polarization SAR speckle suppression method that scattering model and non-local mean are combined | |
CN104463210B (en) | Classification of Polarimetric SAR Image method based on object-oriented and spectral clustering | |
Weissgerber et al. | A temporal estimation of entropy and its comparison with spatial estimations on PolSAR images | |
CN107479052A (en) | Ground concealed target detection method based on Generalized Gaussian Distribution Model | |
CN102323989B (en) | Singular value decomposition non-local mean-based polarized synthetic aperture radar (SAR) data speckle suppression method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180330 Termination date: 20181130 |