CN101329402A - Multi-dimension SAR image edge detection method based on improved Wedgelet - Google Patents
Multi-dimension SAR image edge detection method based on improved Wedgelet Download PDFInfo
- Publication number
- CN101329402A CN101329402A CNA2008101500907A CN200810150090A CN101329402A CN 101329402 A CN101329402 A CN 101329402A CN A2008101500907 A CNA2008101500907 A CN A2008101500907A CN 200810150090 A CN200810150090 A CN 200810150090A CN 101329402 A CN101329402 A CN 101329402A
- Authority
- CN
- China
- Prior art keywords
- wedgelet
- image
- edge
- node
- cost function
- 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
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an SAR image edge detection method based on the multi-scale Wedgelet transformation which relates to the image processing field. The SAR image edge detection method aims at solving the problem that the traditional edge detection method is difficult to suppress the noise and carry out complete and accurate positioning of the edge simultaneously. The procedure is as follows: 1. an SAR image is converted into a gray-scale image which is divided into a plurality of sub-images; 2. the multi-scale Wedgelet decomposition is self-adaptively carried out to each sub-image according to the scale from bottom to top, and the decomposition coefficients of various scales are stored by using a quadtree; 3. the multi-scale Wedgelet approximation is carried out to each sub-image according to the coefficient quadtree obtained by the decomposition, and the multi-scale Wedgelet approximation of the whole image is finally obtained; 4. the edge of the approximated image is detected through edge extraction of gradient operators according to the obtained multi-scale Wedgelet approximated image; 5. the interference of the miscellaneous edges with the length of less than the set threshold on the edge image in the edge image is removed. The SAR image edge detection method has the advantages of fast running time and good edge detection effect, thereby being capable of being used in the edge detection of the SAR image.
Description
Technical field
The invention belongs to technical field of image processing, relate to a kind of method for detecting image edge, can be used for synthetic-aperture radar SAR edge of image is detected.
Background technology
Synthetic-aperture radar SAR is a kind of high-resolution imaging radar, have round-the-clock, multipolarization, from various visual angles, many angles of depression data retrieval capabilities is widely used in military surveillance and civilian remote sensing field.Because imaging radar emission is pure coherent wave, thereby the SAR image is subjected to serious coherent speckle noise influence, and the signal to noise ratio (S/N ratio) of image is lower.Marginal information has great significance for image understanding, and it is the effective way of image segmentation, is again the effective means of feature extraction, develops into the particular study field in Flame Image Process gradually, also is the emphasis of studying in the SAR Flame Image Process.But the existence of SAR image coherent speckle noise has caused difficulty to rim detection.
At present, traditional edge detection method commonly used mainly contains: (1) based on the edge detection method such as the Canny of differentiating operator, Sobel operator, this class operator have advantages such as algorithm is simple, and travelling speed is fast, but very responsive to picture noise.(2) based on the method for test of hypothesis, as the Ratio algorithm, they can overcome The noise preferably, reduce and detect error rate, but edge precision is relatively poor.(3) based on the edge detection method of wavelet theory, the superiority of utilizing Wavelet time-frequency to analyze, the marginal information of detected image under a plurality of yardsticks obtains the image border under the multiscale analysis.But because the SAR image is subjected to the influence of strong coherent speckle noise, detection is difficult to take into account simultaneously squelch and the edge complete and accurate is located traditional algorithm to edge of image, and is unsatisfactory to the detection effect of SAR image.
The uncertainty of rim detection shows that the noise suppression ability of edge detection operator and bearing accuracy are a pair of contradiction, and is though the small scale operator helps the location, edge, very responsive to noise; Though large scale operator noise suppression ability is strong, edge precision is poor, even can lose some local detail.The effective way that provides of head it off is provided in multi-scale morphology, for example, in the multi-scale geometric analysis instrument, the Wedgelet conversion that people such as American scholar Donoho propose has the characteristic of good " line " and " face ", but owing to be subjected to SAR Image Speckle The noise, when this conversion is used for the SAR image, approach image and be easy to generate pseudo-edge, and working time is long, effect and efficient that influence detects.
The content of invention
The objective of the invention is to: overcome in the existing SAR Image Edge-Detection, the very difficult shortcoming of taking into account squelch simultaneously and the edge complete and accurate being located of traditional algorithm, and existing Wedgelet approach the image that obtains have be easy to generate pseudo-edge and working time long shortcoming, propose a kind of SAR method for detecting image edge, make the edge image that obtains after the detection have the complete successional characteristics of good noise immunity and edge based on improvement Wedgelet conversion.
The technical scheme that realizes the object of the invention is: traditional Wedgelet conversion is made improvements, Wedgelet conversion cost function after utilization improves can be caught the characteristic of " line " and " face " in the image well, input picture is carried out multiple dimensioned Wedgelet conversion, the multiple dimensioned Wedgelet that obtains input picture approaches image, carry out rim detection and aftertreatment to approaching image again, remove the interference at tiny assorted edge, obtain final edge image.The specific implementation process is as follows:
(1) input SAR image to be detected is translated into gray level image, and the gray level image that obtains is divided into a plurality of subimages;
(2) with improved Wedgelet cost function each subimage is carried out multiple dimensioned Wedgelet according to bottom-up yardstick adaptively and decompose, preserve the coefficient of dissociation of each yardstick with quaternary tree, the cost function of this Wedgelet conversion is:
Wherein, P ', P " two wedge areas being divided into of the sub-piece S of presentation video respectively,
The size of image block S is n * n,
η is
Weights;
(3) the Wedgelet coefficient that obtains according to decomposition carries out multiple dimensioned Wedgelet reconstruct, and the multiple dimensioned Wedgelet that obtains each subimage approaches, and after all subimages were all approached, the multiple dimensioned Wedgelet that obtains input picture approached;
(4) the multiple dimensioned Wedgelet that obtains is approached image, detect this by the edge extracting gradient operator and approach edge of image;
(5) edge image that obtains of searching and detecting, the length of storing each bar edge is removed the fine edge of edge length less than setting threshold, obtains final edge detection results.
The present invention is owing to adopt multiple dimensioned Wedgelet to approach, and utilize Wedgelet can well catch the characteristic of " line " and " face " in the image, thereby noise immunity is good and edge detection results is complete continuous; Simultaneously, and image is carried out piecemeal handle, further improved its noise immunity and travelling speed, reduced the verification and measurement ratio of pseudo-edge, and made it have the characteristics of applicability because the present invention improves existing multiple dimensioned Wedgelet approach method; Because the present invention has adopted post-processing operation, remove the interference at assorted edge in addition, obtained better rim detection effect.
Description of drawings
Fig. 1 is a schematic flow sheet of the present invention;
Fig. 2 is the Wedgelet dictionary structure figure that the present invention adopts;
Fig. 3 is existing Wedgelet basis function synoptic diagram;
To be the present invention approach synoptic diagram to the Wedgelet of a width of cloth simple image to Fig. 4;
Fig. 5 is that the multiple dimensioned Wedgelet that the present invention adopts approaches Image Edge-Detection gradient operator structural drawing;
Fig. 6 is the figure as a result that the present invention approaches a width of cloth SAR image Wedgelet;
Fig. 7 is that the present invention is used for the SAR image in river and the edge detection results comparison diagram of existing method;
Fig. 8 is the edge detection results comparison diagram that the present invention is used for airport SAR image and existing method;
Fig. 9 is the edge detection results comparison diagram that the present invention is used for airfield runway SAR image and existing method;
Figure 10 is the edge detection results comparison diagram that the present invention is used for harbour SAR image and existing method;
Figure 11 is the figure as a result after the former figure stack of edge detection results of the present invention and SAR image.
Embodiment
With reference to Fig. 1, specific implementation process of the present invention is as follows:
Each subimage size is: M * M, suppose that the input picture size for N * N, makes R=N/M, and then obtain R * R sub-image.
If S is (k
1, k
2, j)={ (x
1, x
2): [k
1/ 2
j, k
1+ 1/2
j] * [k
2/ 2
j, k
2+ 1/2
j] (wherein j 〉=0 is an integer, 0≤k
1, k
2≤ 2
j) be two to advance the set of the point in the square, any two not the line between the summit on same the limit just constituted an edgelet line, the fundamental function on the wedge area that foursquare limit constitutes is advanced at two of each bar edgelet line and place, be called nonsingular Wedgelet, and the fundamental function on square that is not divided then is called the Wedgelet of degeneration.
Step 2 is carried out multiple dimensioned Wedgelet according to bottom-up yardstick with improved Wedgelet cost function adaptively to each subimage and is decomposed, and stores the coefficient of dissociation of each yardstick with quaternary tree.Concrete implementation step is as follows:
1. initialization degree of depth J=log
2The bottom of four fork numbers (M), its corresponding yardstick is j=J;
2. make j=j-1, calculate the cost function of each node of j yardstick, divide following three kinds of situations:
(1) for the leaf node of degenerating, regard its pairing image subblock as a uniform image piece, cost function is:
In the formula,
Presentation video piece S goes up the average of all pixels, the ED-RDP of P presentation video, and the set that promptly approaches the Wedgelet of image, #|P| represents to gather the Wedgelet number among the P, and λ is a complexity punishment parameter;
(2) for nonsingular leaf node, in its corresponding image block, by Wedgelet dictionary structure figure shown in Figure 2, optimum Wedgelet function according to the present invention searches out an optimum edgelet line, and this function is:
In the formula, P ', P " be respectively image subblock S and divided two wedge areas that obtain by an edgelet line, the size of image block S is n * n, (n * n)
2Be used for regulating the requirement of equal value difference shared ratio in the image block of different scale, η is equal value difference part
Weights; By regulating parameter η, can change of the influence of the information of equal value difference part to cost function, η is big more, represents that the information of equal value difference part is big more to the cost function influence, and the present invention gets η=0.3.
Yardstick, position, direction and the fundamental function information of the edgelet line correspondence of the optimum that searches are stored, and calculate the cost function that the optimum Wedgelet of this edgelet line correspondence represents:
With reference to Fig. 3, when j=2, will be on every limit of unit square δ=2 at interval
-jIf a summit has 4 * 2 on the four edges of this unit square
j=16 summits at the two edgelet line left field P ' that advance in the square, have constituted value and have been constant c
aThe Wedgelet basis function, and at the two edgelet line right side area P that advance in the square ", constituted value for constant c
bThe Wedgelet basis function; c
a, c
bValue can obtain by following formula:
c
a=Ave(I(S
j,k)|R
a)
(4)
c
b=Ave(I(S
j,k)|R
b)
In the formula, I two advances square [0,1] for this
2, and
(j, k is presentation video piece S yardstick and position respectively), c
a, c
bBe the average of each pixel in the corresponding Wedgelet zone.
(3), on the yardstick that one deck is thinner down, represent that its cost function is the cost function sum of four child nodes by its 4 child nodes, that is: for intermediate node
costD=costS
1+costS
2+costS
3+costS
4 (5)
Wherein, S
1, S
2, S
3, S
4Four child nodes representing this node correspondence respectively.
3. three kinds of cost function value of each node relatively, get minimum value as the present node cost function value:
cost
optS=min{costE,costP,costD} (6)
4. if cost
OptS=costE or costP, the mark present node is a leaf node, it is all child nodes of root node that beta pruning is fallen with the present node, and the optimum Wedgelet of correspondence position storage present node represents information in decomposing quaternary tree, otherwise the record present node is an intermediate node;
5. (2)~(4) are carried out in circulation, satisfy halt condition j=0, stop after promptly all yardsticks being executed, and obtain the complete coefficient of dissociation quaternary tree of each subimage.
Step 3 according to the Wedgelet coefficient that decomposition obtains, is carried out multiple dimensioned Wedgelet reconstruct, the multiple dimensioned Wedgelet that obtains each subimage approaches, after all subimages were all approached, the multiple dimensioned Wedgelet that obtains input picture approached, and concrete implementation step is as follows:
1. with the coefficient of dissociation quaternary tree, the whole quaternary tree of top-down traversal search;
2. if search leaf node, advance foursquare Wedgelet and represent information according to being stored in its correspondence two in this node, approach the pairing image subblock of this node with Wedgelet;
3. behind complete quaternary tree of traversal search, the image subblock of all leaf node correspondences is approached with Wedgelet, the Wedgelet that finally obtains entire image approaches, as shown in Figure 4.
Step 4 is approached image according to the multiple dimensioned Wedgelet that obtains, and extracts the operator detection by the rim detection gradient and approaches edge of image, and concrete implementation step is as follows:
1. the multiple dimensioned Wedgelet that obtains is approached image, the present invention utilize edge detect gradient formula Y (i, j)=max{Sx, Sy} approaches image to Wedgelet and carries out rim detection, obtain Grad for current pixel point, wherein, the Grad of Sx and Sy corresponding level of difference and vertical direction.The detection template of this formula as shown in Figure 5, among Fig. 5, the pixel in the upper left corner is a current pixel point,
2. according to the edge threshold Th that sets, the Grad of current pixel point is judged, if Y (i, j)>Th, then the value of this pixel is 1, and promptly this pixel is a marginal point, otherwise is 0, and promptly this pixel is non-marginal point.
Step 5 according to detected edge image, is removed length less than the assorted point of the fine edge of a certain threshold value, obtains final edge detection results, and concrete implementation step is as follows:
1. to the resulting edge image of step 4, according to the connected curve in the 8 mode of communicating scanning edge image, every connected curve of mark respectively;
2. add up the length of every connected curve respectively,, in edge image, remove this curve when its length during less than setting threshold Th_post.
Effect of the present invention can further specify by following simulation result.
Simulated conditions:
The present invention has selected for use 4 width of cloth common test SAR images to carry out emulation experiment, and computer hardware is configured to Pentium (R) 4, dominant frequency 3.00GHz.The software platform of existing Wedgelet approximation technique Beamlab200 is Matlab7.0, software platform of the present invention is Matlab7.0 and VC2005, the test pattern of choosing is the Cameraman image, and size 256 * 256 is got the average result of 3 operations working time.
Simulation result:
1. the approach time of approach time of the present invention and prior art is as shown in table 1:
Table 1Beamlab200 tool box and the inventive method contrast working time
Prior art Beamlab200 method (operation platform Matlab7.0) | The inventive method (operation platform Matlab7.0) | The inventive method (operation platform VC2005) | |
Test pattern | The Cameraman image | The Cameraman image | The Cameraman image |
The Wedgelet approach time | 1806 seconds | 178 seconds | 15 seconds |
As seen from Table 1, under the same software platform, the approach time of the present invention only is 1/10 of prior art working time, and under the software platform of VC2005, be lower than 1/100 of prior art working time, as can be seen, compared with prior art, the present invention has more applicability.
2. of the present inventionly approach effect as shown in Figure 6.Wherein, Fig. 6 (a) is the former figure of SAR image, and Fig. 6 (b) is the result that approaches of existing Donoho method, Fig. 6 (c) the inventive method approach the result.From approaching the result as can be seen, the inventive method to approach resultant pseudo-edge less, noise immunity is better.
3. testing result of the present invention such as Fig. 7, Fig. 8, Fig. 9, shown in Figure 10, wherein:
Fig. 7 (a) is the former figure of SAR image in river, Fig. 7 (b) and 7 (b1) are respectively Canny operator aftertreatment front and back edge detection results, Fig. 7 (c) and 7 (c1) are respectively Sobel operator aftertreatment front and back edge detection results, Fig. 7 (d) and 7 (d1) are respectively Ratio algorithm aftertreatment front and back edge detection results, Fig. 7 (e) and 7 (e1) are respectively wavelet modulus maximum method aftertreatment front and back edge detection results, and Fig. 7 (f) and 7 (f1) are respectively the inventive method aftertreatment front and back edge detection results.
Fig. 8 (a) is the former figure of SAR image on airport, Fig. 8 (b) and 8 (b1) are respectively Canny operator aftertreatment front and back edge detection results, Fig. 8 (c) and 8 (c1) are respectively Sobel operator aftertreatment front and back edge detection results, Fig. 8 (d) and 8 (d1) are respectively Ratio algorithm aftertreatment front and back edge detection results, Fig. 8 (e) and 8 (e1) are respectively wavelet modulus maximum method aftertreatment front and back edge detection results, and Fig. 8 (f) and 8 (f1) are respectively the inventive method aftertreatment front and back edge detection results.
Fig. 9 (a) is the former figure of airfield runway SAR image, Fig. 9 (b) and 9 (b1) are respectively Canny operator aftertreatment front and back edge detection results, Fig. 9 (c) and 9 (c1) are respectively Sobel operator aftertreatment front and back edge detection results, Fig. 9 (d) and 9 (d1) are respectively Ratio algorithm aftertreatment front and back edge detection results, Fig. 9 (e) and 9 (e1) are respectively wavelet modulus maximum method aftertreatment front and back edge detection results, and Fig. 9 (f) and 9 (f1) are respectively the inventive method aftertreatment front and back edge detection results.
Figure 10 (a) is the former figure of harbour SAR image, Figure 10 (b) and 10 (b1) are respectively Canny operator aftertreatment front and back edge detection results, Figure 10 (c) and 10 (c1) are respectively Sobel operator aftertreatment front and back edge detection results, Figure 10 (d) and 10 (d1) are respectively Ratio algorithm aftertreatment front and back edge detection results, Figure 10 (e) and 10 (e1) are respectively wavelet modulus maximum method aftertreatment front and back edge detection results, and Figure 10 (f) and 10 (f1) are respectively the inventive method aftertreatment front and back edge detection results.
From Fig. 7, Fig. 8, Fig. 9, Figure 10 testing result as can be seen, it is strong inadequately and the complete continuity in edge is good inadequately that traditional edge detection method suppresses ability to the coherent speckle noise of SAR image, mainly shows as follows:
The detected edge of Canny method is meticulousr, and is more complete to the extraction at edge, but the noise suppression ability is relatively poor, and it is more to be embodied in detected false edges.
Though the detected pseudo-edge of Sobel method is few than the Canny operator, the edge recall rate is not as good as the Canny operator, and the noise suppression ability is undesirable.This mainly is because classical edge detection operator is a Hi-pass filter all in essence, and SAR edge of image and noise all are high-frequency information, detect problem so this class operator can not solve the SAR edge of image well.
The Ratio method has higher recall rate to the SAR image, but detected edge line is thicker, is difficult to differentiate the position of true edge.
Classical relatively gradient operator method of Wavelet Modulus Maxima edge detection method and Ratio method, its noise suppression ability and detect effect and make moderate progress, but to the result of its aftertreatment as can be seen, the many interruption phenomenons of its marginal existence.
The present invention utilizes Wedgelet can catch the characteristic of " line " and " face " in the image preferably, further to the Wedgelet approximate procedure, improved its noise suppression ability and travelling speed, and all had clear improvement aspect noise suppression ability, edge recall rate and the main outline integrality than classic method.
From every kind of method aftertreatment result as can be seen, all there is main outline edge non-continuous event in traditional algorithm, and the inventive method is after post-processing operation, because its good main outline edge continuity, remove the interference at tiny assorted edge, obtained edge detection results preferably.Figure 11 (a), Figure 11 (b), Figure 11 (c), Figure 11 (d) have provided the result of the inventive method to Fig. 7 (a), Fig. 8 (a), Fig. 9 (a), Figure 10 (a) edge detection results and former figure stack respectively, as can be seen from the results, detection is effective to the inventive method to the SAR edge of image.
Claims (5)
1. one kind based on the Multiscale SAR method for detecting image edge that improves Wedgelet, comprises following process:
(1) input SAR image to be detected is translated into gray level image, and the gray level image that obtains is divided into a plurality of subimages;
(2) with improved Wedgelet cost function each subimage is carried out multiple dimensioned Wedgelet according to bottom-up yardstick adaptively and decompose, preserve the coefficient of dissociation of each yardstick with quaternary tree, the cost function of this Wedgelet conversion is:
Wherein, P ', P " two wedge areas being divided into of the sub-piece S of presentation video respectively,
The size of image block S is n * n,
(3) the Wedgelet coefficient that obtains according to decomposition carries out multiple dimensioned Wedgelet reconstruct, and the multiple dimensioned Wedgelet that obtains each subimage approaches, and after all subimages were all approached, the multiple dimensioned Wedgelet that obtains input picture approached;
(4) the multiple dimensioned Wedgelet that obtains is approached image, detect this by the edge extracting gradient operator and approach edge of image;
(5) edge image that obtains of searching and detecting, the length of storing each bar edge is removed the fine edge of edge length less than setting threshold, obtains final edge detection results.
2. Multiscale SAR method for detecting image edge according to claim 1, wherein step (2) is carried out according to the following procedure:
(2a) according to the cost function calculation formula
Set up the bottom of coefficient of dissociation quaternary tree, obtain bottom coefficient,
Wherein, the ED-RDP of P presentation video, the set that promptly approaches the Wedgelet of image,
λ is a complexity punishment parameter;
(2b) searching on each yardstick each respectively by bottom-up yardstick two advances foursquare optimum Wedgelet and represents, this optimum represents to be divided into that non degenerate Wedgelet represents, degeneration Wedgelet represents and this two advances square for three kinds of situations of intermediate node, represent that for nonsingular Wedgelet the cost function of searching for its optimum Wedgelet base is:
Represent that for the Wedgelet that degenerates its cost function is:
For this two advance the square be the situation of intermediate node, cost function be its correspondence four child nodes the cost function sum;
(2c) reckling of getting in described three cost function value advances the cost function of square corresponding node as working as the first two, is stored in the node of coefficient of dissociation quaternary tree correspondence;
If (2d) corresponding degeneration of minimum cost function or nonsingular Wedgelet represent, the mark present node is a leaf node, it is all child nodes of root node that beta pruning is fallen with the present node, and the optimum Wedgelet of the correspondence position in coefficient of dissociation quaternary tree storage present node represents information, otherwise the record present node is an intermediate node;
(2e) (2b)~(2d) carried out in circulation, stops after all yardsticks are executed, and obtains the complete coefficient of dissociation quaternary tree of each subimage.
3. Multiscale SAR method for detecting image edge according to claim 1, wherein step (3) is carried out according to the following procedure:
(3a) with the coefficient of dissociation quaternary tree, the whole quaternary tree of top-down traversal search;
(3b) for searching each leaf node, advance foursquare Wedgelet and represent information according to being stored in its correspondence two in this node, approach the pairing image subblock of this node with Wedgelet;
(3c) behind complete quaternary tree of traversal search, then the image subblock of all leaf node correspondences approaches with Wedgelet, and the Wedgelet that finally obtains entire image approaches.
4. Multiscale SAR method for detecting image edge according to claim 1, wherein step (4) is carried out according to the following procedure:
(4a) approach the characteristics of image according to multiple dimensioned Wedgelet, utilize rim detection gradient formula Y (i, j)=max{Sx, Sy} approaches image to Wedgelet and carries out rim detection, obtain the Grad of current pixel point, wherein, the Grad of Sx and Sy corresponding level of difference and vertical direction;
(4b), the Grad of current pixel point is judged according to the edge threshold Th that sets, if Y (i, j)>Th, then the value of this pixel is 1, and promptly this pixel is a marginal point, otherwise is 0, and promptly this pixel is non-marginal point.
5. Multiscale SAR method for detecting image edge according to claim 1, wherein step (5) is carried out according to the following procedure:
(1). to the resulting edge image of step 4, according to the connected curve in the 8 mode of communicating scanning edge image, every connected curve of mark respectively;
(2). add up the length of every connected curve respectively,, in edge image, remove this curve when its length during less than setting threshold Th_post.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008101500907A CN101329402B (en) | 2008-06-20 | 2008-06-20 | Multi-dimension SAR image edge detection method based on improved Wedgelet |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008101500907A CN101329402B (en) | 2008-06-20 | 2008-06-20 | Multi-dimension SAR image edge detection method based on improved Wedgelet |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101329402A true CN101329402A (en) | 2008-12-24 |
CN101329402B CN101329402B (en) | 2010-12-29 |
Family
ID=40205303
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2008101500907A Expired - Fee Related CN101329402B (en) | 2008-06-20 | 2008-06-20 | Multi-dimension SAR image edge detection method based on improved Wedgelet |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101329402B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102122386A (en) * | 2011-03-01 | 2011-07-13 | 西安电子科技大学 | SAR (stop and reveres) image segmentation method based on dictionary migration clustering |
CN102156971A (en) * | 2011-04-15 | 2011-08-17 | 西安电子科技大学 | Speckle suppression method of synthetic aperture radar (SAR) image based on linear singularity information |
CN101777181B (en) * | 2010-01-15 | 2012-01-04 | 西安电子科技大学 | Ridgelet bi-frame system-based SAR image airfield runway extraction method |
CN102325252A (en) * | 2011-08-08 | 2012-01-18 | 哈尔滨工业大学深圳研究生院 | Optimal-wavelet-packet-based joint source channel coding method and image transmission system |
CN104966296A (en) * | 2015-06-16 | 2015-10-07 | 西南石油大学 | Moving window N-Smoothlets image edge detection method |
CN105023257A (en) * | 2015-08-24 | 2015-11-04 | 西南石油大学 | N-Smoothlets based image denoising method |
CN107817493A (en) * | 2017-10-25 | 2018-03-20 | 电子科技大学 | A kind of 2D near fields diameter radar image method for reconstructing based on compressed sensing |
CN108399654A (en) * | 2018-02-06 | 2018-08-14 | 北京市商汤科技开发有限公司 | It retouches in the generation of special efficacy program file packet and special efficacy generation method and device when retouching |
CN108562899A (en) * | 2018-04-20 | 2018-09-21 | 西安电子科技大学 | High-resolution polarimetric SAR target image rapid simulation method |
CN110363212A (en) * | 2019-03-28 | 2019-10-22 | 西南石油大学 | More wedgelet image approximate methods based on boundary scan |
CN111208820A (en) * | 2020-01-09 | 2020-05-29 | 哈尔滨工程大学 | Particle unmanned vehicle group under artificial intelligence big data, control method and medium |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101126811A (en) * | 2007-09-29 | 2008-02-20 | 北京交通大学 | Method for detecting lakeshore and extracting lake profile from SAR image |
-
2008
- 2008-06-20 CN CN2008101500907A patent/CN101329402B/en not_active Expired - Fee Related
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101777181B (en) * | 2010-01-15 | 2012-01-04 | 西安电子科技大学 | Ridgelet bi-frame system-based SAR image airfield runway extraction method |
CN102122386B (en) * | 2011-03-01 | 2012-08-08 | 西安电子科技大学 | SAR (stop and reveres) image segmentation method based on dictionary migration clustering |
CN102122386A (en) * | 2011-03-01 | 2011-07-13 | 西安电子科技大学 | SAR (stop and reveres) image segmentation method based on dictionary migration clustering |
CN102156971A (en) * | 2011-04-15 | 2011-08-17 | 西安电子科技大学 | Speckle suppression method of synthetic aperture radar (SAR) image based on linear singularity information |
CN102325252A (en) * | 2011-08-08 | 2012-01-18 | 哈尔滨工业大学深圳研究生院 | Optimal-wavelet-packet-based joint source channel coding method and image transmission system |
CN104966296B (en) * | 2015-06-16 | 2018-01-09 | 西南石油大学 | Sliding window N Smoothlets method for detecting image edge |
CN104966296A (en) * | 2015-06-16 | 2015-10-07 | 西南石油大学 | Moving window N-Smoothlets image edge detection method |
CN105023257B (en) * | 2015-08-24 | 2018-01-30 | 西南石油大学 | Image de-noising method based on N Smoothlets |
CN105023257A (en) * | 2015-08-24 | 2015-11-04 | 西南石油大学 | N-Smoothlets based image denoising method |
CN107817493A (en) * | 2017-10-25 | 2018-03-20 | 电子科技大学 | A kind of 2D near fields diameter radar image method for reconstructing based on compressed sensing |
CN107817493B (en) * | 2017-10-25 | 2021-05-14 | 电子科技大学 | 2D near-field synthetic aperture radar image reconstruction method based on compressed sensing |
CN108399654A (en) * | 2018-02-06 | 2018-08-14 | 北京市商汤科技开发有限公司 | It retouches in the generation of special efficacy program file packet and special efficacy generation method and device when retouching |
US11640683B2 (en) | 2018-02-06 | 2023-05-02 | Beijing Sensetime Technology Development Co., Ltd. | Stroke special effect program file package generating method and apparatus, and stroke special effect generating method and apparatus |
CN108562899A (en) * | 2018-04-20 | 2018-09-21 | 西安电子科技大学 | High-resolution polarimetric SAR target image rapid simulation method |
CN108562899B (en) * | 2018-04-20 | 2021-08-20 | 西安电子科技大学 | High-resolution polarization SAR target image rapid simulation method |
CN110363212A (en) * | 2019-03-28 | 2019-10-22 | 西南石油大学 | More wedgelet image approximate methods based on boundary scan |
CN111208820A (en) * | 2020-01-09 | 2020-05-29 | 哈尔滨工程大学 | Particle unmanned vehicle group under artificial intelligence big data, control method and medium |
Also Published As
Publication number | Publication date |
---|---|
CN101329402B (en) | 2010-12-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101329402B (en) | Multi-dimension SAR image edge detection method based on improved Wedgelet | |
CN105608691B (en) | A kind of High-resolution SAR Images single building extracting method | |
CN104049245B (en) | Urban building change detection method based on LiDAR point cloud spatial difference analysis | |
CN103500338B (en) | Road zebra crossing extraction method based on Vehicle-borne Laser Scanning point cloud | |
Sithole et al. | Comparison of filtering algorithms | |
CN104008553B (en) | Crack detection method with image gradient information and watershed method conflated | |
CN101551456B (en) | Method for detecting water area margin of SAR image based on improved shearlet transformation | |
CN101996401B (en) | Target analysis method and apparatus based on intensity image and depth image | |
CN103208123B (en) | Image partition method and system | |
CN101950364A (en) | Remote sensing image change detection method based on neighbourhood similarity and threshold segmentation | |
CN106778668B (en) | A kind of method for detecting lane lines of robust that combining RANSAC and CNN | |
CN102567994B (en) | Infrared small target detection method based on angular point gaussian characteristic analysis | |
CN102592128B (en) | Method and device for detecting and processing dynamic image and display terminal | |
CN101833664A (en) | Video image character detecting method based on sparse expression | |
CN107507170A (en) | A kind of airfield runway crack detection method based on multi-scale image information fusion | |
CN103048329A (en) | Pavement crack detecting method based on active contour model | |
CN103729842A (en) | Fabric defect detection method based on local statistical characteristics and overall significance analysis | |
CN107450054B (en) | A kind of adaptive Coherent Noise in GPR Record denoising method | |
CN104200471A (en) | SAR image change detection method based on adaptive weight image fusion | |
CN105741276A (en) | Ship waterline extraction method | |
CN108446637B (en) | SAR image change detection method based on stereogram model | |
Rajeswari et al. | Automatic road extraction based on level set, normalized cuts and mean shift methods | |
Wang et al. | A method for detecting windows from mobile LiDAR data | |
Zhang et al. | A framework for automated construction of building models from airborne Lidar measurements | |
CN112198170A (en) | Detection method for identifying water drops in three-dimensional detection of outer surface of seamless steel pipe |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20210402 Address after: 710076 room 104, block B2, software new town phase II, tianguba Road, Yuhua Street office, high tech Zone, Xi'an City, Shaanxi Province Patentee after: Discovery Turing Technology (Xi'an) Co.,Ltd. Address before: 710071 No. 2 Taibai Road, Shaanxi, Xi'an Patentee before: XIDIAN University |
|
TR01 | Transfer of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20101229 |
|
CF01 | Termination of patent right due to non-payment of annual fee |