CN106504263B - A kind of quick continuous boundary extracting method of image - Google Patents
A kind of quick continuous boundary extracting method of image Download PDFInfo
- Publication number
- CN106504263B CN106504263B CN201610963593.0A CN201610963593A CN106504263B CN 106504263 B CN106504263 B CN 106504263B CN 201610963593 A CN201610963593 A CN 201610963593A CN 106504263 B CN106504263 B CN 106504263B
- Authority
- CN
- China
- Prior art keywords
- image
- result
- edge
- data matrix
- value
- 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 present invention provides a kind of quick continuous boundary extracting method of image, comprising: carries out edge detection to image I using differential detection operator, obtains edge detection results;Skeletal extraction is carried out to edge detection results, obtains skeletal extraction result;Each pixel in skeletal extraction result is traversed, if current pixel point value is not 0, the quantity that pixel point value in the nine grids centered on current pixel point is not 0 is counted, marks current pixel point if quantity is 2, be put into breakpoint and extract results set;Continuation breakpoint extracts as a result, obtaining continuation result;Continuation result is expanded to width r;Generate the imbedding function primary data matrix of partial differential equation;The continuous boundary that partial differential equation obtain image is iteratively solved, image continuous boundary is obtained and extracts result.The present invention can obtain the good edge of continuity during image processing, with accuracys such as the image segmentation, measurement, the matchings that preferably improve the later period.
Description
Technical field
The invention belongs to technical field of image processing, and in particular to a kind of quick continuous boundary extracting method of image.
Background technique
In existing edge detection method, method of differential operator and level set method are the representational methods of two classes: differential
Detective operators method is with Canny operator (John Canny, ' A computational approach to edge
Detection ', IEEE Transactions on Pattern Analysis and Machine Intelligence,
1986,8, (6), pp.679-698) it is representative, the advantages of such methods is that detection speed is fast, the disadvantage is that edge detection results packet
Containing " fracture " part;The Typical Representative of level set method includes MCM (Sapiro G.Affine invariant scale space
[J] .International Journal of Computer Vision, 1993,11 (1): 25-44.) etc., such methods
Advantage is that edge extracting result is continuously good, and it is related to the position of the equal initial profile of testing result that disadvantage executes speed.In recent years, all
More scholars be still directed to above-mentioned respective problem improved (Xiaohu Lu, Jian Yaoyao, Kai Li, Li Li,
' Cannylines:A parameter-free line segment detector ', IEEE International
Conference on Image Processing, 2015, pp.507-511, Di Jia, Cheng-long Xiao, Jin-
Guang Sun, ' Edge detection method of gaussian block distance ', IEEE
International Conference on Image Processing, 2015, pp.3049-3053, Marcelo
Pereyra, Hadj Batatia, Steve McLaughlin. ' Exploiting Information Geometry to
Improve the Convergence of Nonparametric Active Contours ', IEEE Transactions
On image processing, 2015,24, (3), pp.836-845), to obtain ideal edge extracting result.However,
These improvement fail fundamentally to solve the disadvantage that two class methods, in the recent period, a kind of continuity edge extracting method of image (merchant
Enlightening, Dong Na, Meng Xiangfu, Li Sihui computer engineering and science .2015,37 (2): 384-389) provide it is excellent in conjunction with two class methods
A kind of rapid succession edge extracting method of point, this method use template detection fracture position, and on the position with certain
Radius carry out round expansion, when the edge of fracture is closer to, dilation can connect, recycling MCM model with
Expansion plans are that initial position is iterated solution, obtain continuous boundary.Although preferable experimental result can be obtained, still deposit
In following problems: when 1) carrying out circle expansion with radius r, when two crack edge breakpoint hypertelorisms, the value of r just needs to increase
The r value for adding, and increasing will lead to the edge iterative solution mistake that other breakpoints are closer;2) MCM is using curvature as driving force
Edge-stopping function, do not consider edge gradient stop force and tractive force, image true edge cannot be obtained well.Namely
It says, both of these problems limit its specific use environment, to reduce practicability.
Summary of the invention
In view of the problems of the existing technology, the present invention provides a kind of image continuous boundary extracting method.
Technical scheme is as follows:
A kind of quick continuous boundary extracting method of image, comprising the following steps:
Step 1 carries out edge detection to image I using differential detection operator, obtains edge detection results E;
Step 2 carries out skeletal extraction to edge detection results E, obtains skeletal extraction result Ec;
Each pixel does following processing in step 3, traversal skeletal extraction result Ec, obtains breakpoint and extracts result B: if
Current pixel point value is not 0, counts the quantity that pixel point value in the nine grids centered on current pixel point is not 0, if number
Amount marks current pixel point for 2, is put into breakpoint and extracts in results set;Result B is extracted according to setting step-length s continuation breakpoint,
I.e. centered on current pixel point, rectilinear direction progress is formed by along two pixels that breakpoint extracts in result B not for 0
The pixel that s pixel value is 0 in the rectilinear direction is set 1 by continuation;Finally obtain continuation result Ee;
Continuation result Ee is expanded to width r by step 4, obtains expansion results P;R is greater than 1 positive integer;
Step 5, based on expansion results P, generate the imbedding function primary data matrix u of partial differential equation0;
Step 6, iterative solution partial differential equation obtain the continuous boundary of image, obtain image continuous boundary and extract result.
The step 5 is specifically: if the i-th row jth column pixel value is 0 in expansion results P, inserting for the 0th moment for 255
The data matrix u of imbedding function u0The i-th row jth column, i.e., the primary data matrix of imbedding function the i-th row jth columnIf
The i-th row jth column pixel value is greater than 0 in expansion results P, then by the data matrix u of -255 the 0th moment imbedding function u of filling0?
I row jth column, i.e. the i-th row jth column of the primary data matrix of imbedding functionIt traverses in expansion results P as procedure described above
All pixels obtain the primary data matrix u of imbedding function u0。
The step 6 is specifically using following improved MCM partial differential equation:
Wherein,Partial derivative is asked to the time for curve C, g is Edge-stopping function, For image I's
Gradient value,K is the curvature of image I,N is the gradient normal vector of image I, and K is choosing
Fixed constant, for controlling the fall off rate of g, K value is bigger, thenByInfluence it is smaller;IxxExist for image I
Second dervative on the direction x, IyyFor the second dervative of image I in y-direction, IxyAfter first seeking local derviation in the direction x to image I,
Partial derivative is reversely sought in y again.
The improved MCM partial differential equation obtain iterative solution equation using discretization method:
Wherein,For in the data matrix of n+1 moment imbedding function u the i-th row jth arrange value,Letter is embedded in for the n moment
The value of i-th row jth column in the data matrix of number u, Δ t are iteration step length,For the derivative of g;In primary iteration,Value
ForAt the end of iterative solution,For last iteration result, i.e., the point in the data matrix of imbedding function u not for 0 is figure
The edge of picture.
The utility model has the advantages that
The present invention carries out edge detection to image using differential detection operator, obtains edge detection results;To edge detection
As a result skeletal extraction is carried out, skeletal extraction result is obtained;Each pixel in skeletal extraction result is traversed, if current pixel point
Value is not 0, counts the quantity that pixel point value in the nine grids centered on current pixel point is not 0, is marked if quantity is 2
Current pixel point extracts result as breakpoint;It extracts according to step-length continuation breakpoint is set as a result, i.e. with current pixel point in
The heart is formed by rectilinear direction progress continuation along two pixels that breakpoint extracts in result not for 0, will be in the rectilinear direction
The pixel that s pixel value is 0 sets 1, obtains continuation result;Continuation result is expanded to width r, obtains expansion results;With expansion
As a result based on, the imbedding function initial value of partial differential equation is generated;Iteratively solve the continuous side that partial differential equation obtain image
Edge obtains image continuous boundary and extracts result.The present invention can obtain the good edge of continuity during image processing, with more
The accuracys such as image segmentation, measurement, the matching in later period are improved well.The fast advantage of the differential detection operator processing speed utilized,
It extracts breakpoint location and expands edge using the connection of the characteristic of edge nature continuation, and using the result as the initial of imbedding function
Value is obtained using the iterative solution of continuity Boundary extracting algorithm in -255 to+255 tonal range and finally extracts result.Cause
Edge graph very close true edge is expanded for connection, therefore iteratively solves process quickly, is obtained with this quickly continuous
Edge extracting result.
Detailed description of the invention
Fig. 1 is a kind of flow chart of the quick continuous boundary extracting method of image in the specific embodiment of the invention;
Fig. 2 a~h is 8 kinds of nine grids in the specific embodiment of the invention respectively;
Fig. 3 is improved MCM partial differential equation schematic diagram in the specific embodiment of the invention;
Fig. 4 a~h is the process schematic that image continuous boundary extracts in the specific embodiment of the invention, and a is one to be checked
The line segment of survey, b are to carry out edge detection using differential detection operator Canny to obtain edge detection results;D is by C and E construction
Vector ray, e are the vector ray progress continuation acquisition along C and E construction as a result, the crack edge connection result that f is figure a, g
For the expansion plans of f, h is the border seal curve of g;
Fig. 5 a~j be in the specific embodiment of the invention scene image continuous boundary extract as a result, a be original image, b~
D is respectively the edge detection experimental result for using document A, document B, Canny operator to obtain, and e is the side obtained using step 1-4
As a result, f is the edge of iterative solution, g~j is respectively the partial enlarged view of b, d, e, f for edge continuation;
Fig. 6 a~j is to build object image continuous boundary in the specific embodiment of the invention to extract as a result, a is original image, b
~d is respectively the edge detection experimental result for using document A, document B, Canny operator to obtain, and e is to be obtained using step 1-4
For edge extension as a result, f is the edge of iterative solution, g~j is respectively the partial enlarged view of b, d, e, f;
Fig. 7 a~j is that the extraction of object image continuous boundary is let people in the specific embodiment of the invention as a result, a is original image, b
~d is respectively the edge detection experimental result for using document A, document B, Canny operator to obtain, and e is to be obtained using step 1-4
For edge extension as a result, f is the edge of iterative solution, g~j is respectively the partial enlarged view of b, d, e, f.
Specific embodiment
Specific embodiments of the present invention will be described in detail with reference to the accompanying drawing.
A kind of quick continuous boundary extracting method of image as shown in Figure 1, comprising the following steps:
Step 1 carries out edge detection to image I using differential detection operator Canny, obtains edge detection results E;
Step 2 carries out skeletal extraction using K3M algorithm to edge detection results E, obtains skeletal extraction result Ec;
Each pixel does following processing in step 3, traversal skeletal extraction result Ec, obtains breakpoint and extracts result B: if
Current pixel point value is not 0, counts the quantity that pixel point value in the nine grids centered on current pixel point is not 0, if number
Amount marks current pixel point for 2, is put into breakpoint and extracts in results set;Result B is extracted according to setting step-length s continuation breakpoint,
I.e. centered on current pixel point, rectilinear direction progress is formed by along two pixels that breakpoint extracts in result B not for 0
The pixel that s pixel value is 0 in the rectilinear direction is set 1 by continuation;Finally obtain continuation result Ee;
The nine grids be divided into " upper left ", on ", " upper right ", " left ", " in ", " right side in ", " lower-left ", "lower", " right
Under ", in present embodiment, current pixel point value is not that pixel point value is not 0 in 0 and the nine grids centered on current pixel point
Quantity be 2 the case where, the nine grids of 8 kinds of situations as shown in Fig. 2 a~h, the possessive case set 0 and will " in " set 1, then in addition to
" in " lattice, successively set in 8 nine grids respectively by said sequence 1 to get to 8 kinds of nine grids in each of there are 2 lattice to be
1。
Continuation result Ee is expanded to width r by step 4, obtains expansion results P;R is greater than 1 positive integer;
Step 5, based on expansion results P, generate the imbedding function primary data matrix u of partial differential equation0;
If the i-th row jth column pixel value P in expansion results PijIt is 0, then by the number of 255 the 0th moment imbedding function u of filling
According to matrix, i.e. the i-th row jth column of the primary data matrix of imbedding functionIf the i-th row jth column pixel value in expansion results P
PijGreater than 0, then by the data matrix of -255 the 0th moment imbedding function u of filling, i.e. the i-th of the primary data matrix of imbedding function
Row jth columnThe all pixels in expansion results P are traversed as procedure described above, obtain the primary data matrix u of imbedding function0;
I-th row jth of the imbedding function primary data matrix of partial differential equation arranges
Wherein, m, n are respectively the length and width of image I.
Step 6, iterative solution partial differential equation obtain the continuous boundary of image, obtain image continuous boundary and extract result.
MCM equation is a kind of equation based on linear geometry hot-fluid theory, and evolutionary process keeps closure and the company of curve
The general character:
Classical MCM equation evolutionary process does not account for the resistance problem of object edge.At a kind of " continuity edge of image
Extracting method " in the problem is improved, increase the Edge-stopping function g using curvature as driving force, define shape
Formula is as follows:
Wherein, g is Edge-stopping function,F is curvature item, Ix、IyFor
First differential, Ixx、IyyFor second-order differential.By adjusting contrast parameter K, different curvature degrees of regulation can be obtained, to control
The decrease speed of g.The problem of classical MCM equation be it is still poor for higher curvature marginal portion drag effect, be this this reality
It applies in mode using following improved MCM partial differential equation:
Wherein,Partial derivative is asked to the time for curve C, g is Edge-stopping function, For image I's
Gradient value,K is the curvature of image I,N is the gradient normal vector of image I, and K is choosing
Fixed constant, for controlling the fall off rate of g, K value is bigger, thenByInfluence it is smaller;IxxExist for image I
Second dervative on the direction x, IyyFor the second dervative of image I in y-direction, IxyAfter first seeking local derviation in the direction x to image I,
Partial derivative is reversely sought in y again.Edge-stopping function g is the function about image gradient, gradientBigger, then g is smaller, more leans on
Nearly image border drives evolution strength smaller at this time;Meanwhile it introducingIncrease edge suction, principle is as shown in figure 3, figure
In closed curve indicate " object " edge, it is assumed that interior intensity value is relatively external low.Since gradient modulus value reaches at edge
Local maximum, therefore Edge-stopping function g reaches local minimum.It uses respectivelyAndCurve indicates.Due toAlways
Be directed toward g increase direction, no matter therefore interior of articles or outside,Always point at the direction at offline edge.It is assumed that curve C
(t) near east to object edge, curvilinear inner is always pointed at by the normal direction N of regulation C (t).Therefore, if current curves
Position is in border outer, then N will be withIt is contrary, i.e.,For negative value, thereforeIt is consistent with normal orientation,
This effect at this time is to move C (t) outside outer boundary to closer to boundary direction;, whereas if current curves position is
Inside edge, then since N will be withDirection is consistent, i.e.,For positive value, thereforeWith normal orientation on the contrary, i.e.
Effect be move C (t) from border inner to closer to boundary direction, thereforeIncrease edge attraction.
Using Level Set Method, since there are following conversion formulas:
Then formulaThe partial differential equation of corresponding imbedding function u are as follows:
Improved MCM partial differential equation obtain iterative solution equation using discretization method:
Wherein,For in the data matrix of n+1 moment imbedding function u the i-th row jth arrange value,Letter is embedded in for the n moment
The value of i-th row jth column in the data matrix of number u, Δ t are iteration step length,For the derivative of g;In primary iteration,Value
ForAt the end of iterative solution,For last iteration result, i.e., the point in the data matrix of imbedding function u not for 0 is figure
The edge of picture.
Fig. 4 gives the process schematic that the quick continuous boundary of image in present embodiment extracts: a is one to be detected
Line segment, b be using differential detection operator Canny carry out edge detection obtain edge detection results, E is the breakpoint bit detected
It sets, C is the neighborhood territory pixel point of E, and these two types point is obtained by step 3;D is by the vector ray of C and E construction, and e is along C and E structure
It is that the vector ray made carries out continuation acquisition that g is the expansion plans of f as a result, f be the crack edge connection result for scheming a, h for g side
Edge closed curve, as seen from the figure, the curve will be in true edge closing and curves.Construction improves MCM model based on scheming g
Initial curve, and final object edge is solved by the contraction of curve.
Since this method has made iterative edge be located at original image adjacent edges, can be completed by less secondary iteration
Edge extracting simultaneously because improved MCM partial differential equation have from merging stalling characteristic, therefore can merge therefore can close
And the trifling region in distance map, to ensure that the scale of image segmentation with preferable granularity.
Experimental analysis: CPU frequency 3.2GHz is used, memory 2G is as experimental situation, MATLAB programming.In order to preferably say
The validity of bright the method for the present invention, using image different classes of in natural scene, building, personage three as experimental data,
Fig. 5~Fig. 7 gives one group of experimental result.In Fig. 5~Fig. 7, a is original image, b~d be respectively use document A (Di Jia,
Cheng-long Xiao, Jin-guang Sun, ' Edge detection method of gaussian block
Distance ', IEEE International Conference on Image Processing, 2015, pp.3049-
3053), document B (Marcelo Pereyra, Hadj Batatia, Steve McLaughlin, ' Exploiting
Information Geometry to Improve the Convergence of Nonparametric Active
Contours ', IEEE Transactions on image processing, 2015,24, (3), pp.836-845), Canny
The edge detection experimental result that operator obtains.From experimental result, the careful degree at b and d detection edge is better than c, but the side of c
Edge continuity is good, such as the beautiful and text in the flower in Fig. 5, the building roof in Fig. 6, Fig. 7.E is to be obtained using step 1-4
Edge extension as a result, as seen from the figure, crack edge is attached.
E and d are as it can be seen that e has preferably repaired the edge breaks part in d, although with step in 5~Fig. 7 of comparison diagram
This method of rapid 1-4 may generate false edge, but be iteratively solved by improved MCM partial differential equation, repeatedly
Automatic calculation obtains the true edge of image during generation.Using e as input, 50 acquisition f of iteration, it can be clearly seen that, f is not
It has only preferably handled by repairing pseudo-edge, and the continuity at extracted edge is good.G~j is respectively that the part of b, d, e, f are put
Big figure, marked region that observation j is provided as it can be seen that the edge continuity that obtains of the method for the present invention more preferably, such as g in Fig. 5 and i
Blade-section poor continuity, and this partial region is not detected in the h in Fig. 5;Black roof in Fig. 6, the arm in Fig. 7 and
Body marginal portion all there is a problem of identical.
Table 1 gives the processing time correlation data of edge detection.With the reduction of picture size, all edge detections
It is on a declining curve to handle the time.Wherein, the processing speed of Canny operator is most fast, and the processing speed of document A method is higher than
Canny method.Document B is a kind of method based on level set class, needs to iteratively solve, therefore speed is slower than first two method.
Since the processing time of the method for the present invention is based on Canny differential operator, the processing time is slightly longer than Canny method, but
The time is handled compared with method in document A and document B will less very much, and respectively less than 1 second.
Table 1 handles time comparison (second)
Picture size | Canny | Document A | Document B | This method | |
Natural scene | 290*290 | 0.08 | 0.59 | 5.63 | 0.32 |
Building | 320*240 | 0.078 | 0.56 | 5.2 | 0.23 |
Personage | 240*240 | 0.056 | 0.52 | 3.6 | 0.12 |
Claims (3)
1. a kind of image continuous boundary extracting method, which comprises the following steps:
Step 1 carries out edge detection to image I using differential detection operator, obtains edge detection results E;
Step 2 carries out skeletal extraction to edge detection results E, obtains skeletal extraction result Ec;
Each pixel does following processing in step 3, traversal skeletal extraction result Ec, obtains breakpoint and extracts result B: if current
Pixel point value is not 0, counts the quantity that pixel point value in the nine grids centered on current pixel point is not 0, if quantity is 2
Current pixel point is then marked, breakpoint is put into and extracts in results set;Result B is extracted according to setting step-length s continuation breakpoint, i.e., to work as
Centered on preceding pixel point, it is formed by rectilinear direction progress continuation along two pixels that breakpoint extracts in result B not for 0,
The pixel that s pixel value is 0 in the rectilinear direction is set 1;Finally obtain continuation result Ee;
Continuation result Ee is expanded to width r by step 4, obtains expansion results P;
Step 5, based on expansion results P, generate the imbedding function primary data matrix u of partial differential equation0;
Step 6, iterative solution partial differential equation obtain the continuous boundary of image, obtain image continuous boundary and extract result;
The step 6 is specifically using following improved MCM partial differential equation:
Wherein,Partial derivative is asked to the time for curve C,gFor Edge-stopping function, For the gradient of image I
Value,K is the curvature of image I,N is the gradient normal vector of image I, and K is selected
Constant, for controlling the fall off rate of g, Ix、IyFor first differential;IxxFor the second dervative of image I in the x direction, IyyFor figure
As the second dervative of I in y-direction, IxyAfter first seeking local derviation in the direction x to image I, then in the direction y seek partial derivative.
2. the method according to claim 1, wherein the step 5 is specifically: if the i-th row in expansion results P
Jth column pixel value is 0, then by the data matrix u of 255 the 0th moment imbedding function u of filling0The i-th row jth column, i.e. imbedding function
Primary data matrix the i-th row jth columnIf the i-th row jth column pixel value is greater than 0 in expansion results P, -255 are filled out
Enter the data matrix u of the 0th moment imbedding function u0The i-th row jth column, i.e. the i-th row jth of the primary data matrix of imbedding function
ColumnThe all pixels in expansion results P are traversed as procedure described above, obtain the primary data matrix u of imbedding function u0。
3. the method according to claim 1, wherein the improved MCM partial differential equation use discretization side
Method obtains iterative solution equation:
Wherein,For in the data matrix of n+1 moment imbedding function u the i-th row jth arrange value,For n moment imbedding function u's
The value that the i-th row jth arranges in data matrix, △ t are iteration step length,For the derivative of g;In primary iteration,Value beAt the end of iterative solution,For last iteration result, i.e., the point in the data matrix of imbedding function u not for 0 is image
Edge.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610963593.0A CN106504263B (en) | 2016-11-04 | 2016-11-04 | A kind of quick continuous boundary extracting method of image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610963593.0A CN106504263B (en) | 2016-11-04 | 2016-11-04 | A kind of quick continuous boundary extracting method of image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106504263A CN106504263A (en) | 2017-03-15 |
CN106504263B true CN106504263B (en) | 2019-07-12 |
Family
ID=58321704
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610963593.0A Expired - Fee Related CN106504263B (en) | 2016-11-04 | 2016-11-04 | A kind of quick continuous boundary extracting method of image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106504263B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109685074B (en) * | 2018-10-17 | 2022-05-10 | 福州大学 | Bank card number row positioning method based on Scharr operator |
CN109544577B (en) * | 2018-11-27 | 2022-10-14 | 辽宁工程技术大学 | Improved straight line extraction method based on edge point grouping |
CN111311573B (en) * | 2020-02-12 | 2024-01-30 | 贵州理工学院 | Branch determination method and device and electronic equipment |
CN115880699B (en) * | 2023-03-03 | 2023-05-09 | 济南市莱芜区综合检验检测中心 | Food packaging bag detection method and system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102270299A (en) * | 2011-08-24 | 2011-12-07 | 复旦大学 | Edge connection algorithm realized in parallel based on breakpoints |
CN103530878A (en) * | 2013-10-12 | 2014-01-22 | 北京工业大学 | Edge extraction method based on fusion strategy |
CN103824268A (en) * | 2014-02-08 | 2014-05-28 | 江西赛维Ldk太阳能高科技有限公司 | Crystal grain image edge connecting method and apparatus |
-
2016
- 2016-11-04 CN CN201610963593.0A patent/CN106504263B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102270299A (en) * | 2011-08-24 | 2011-12-07 | 复旦大学 | Edge connection algorithm realized in parallel based on breakpoints |
CN103530878A (en) * | 2013-10-12 | 2014-01-22 | 北京工业大学 | Edge extraction method based on fusion strategy |
CN103824268A (en) * | 2014-02-08 | 2014-05-28 | 江西赛维Ldk太阳能高科技有限公司 | Crystal grain image edge connecting method and apparatus |
Non-Patent Citations (4)
Title |
---|
一种图像的连续性边缘提取方法;贾 迪等;《计算机工程与科学》;20150228;第37卷(第2期);第384-387页 * |
一种基于断点处边缘方向保持假设的闭合轮廓提取方法;关涛等;《计算机学报》;20140630;第37卷(第6期);第1335-1341页 * |
基于主动生长的断裂裂缝块的连接方法;朱平哲等;《计算机应用》;20111231;第31卷(第12 期);第3382-3384页 * |
基于边缘吸引力场正则化的短程线主动轮廓模型;岑 峰等;《电子学报》;20030131;第31卷(第1期);第17-18页 * |
Also Published As
Publication number | Publication date |
---|---|
CN106504263A (en) | 2017-03-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106504263B (en) | A kind of quick continuous boundary extracting method of image | |
Miller et al. | Geodesic shooting for computational anatomy | |
Shi et al. | A fast level set method without solving pdes [image segmentation applications] | |
CN104392241B (en) | A kind of head pose estimation method returned based on mixing | |
CN101571951B (en) | Method for dividing level set image based on characteristics of neighborhood probability density function | |
CN105976364B (en) | Based on the statistical average model building method for simplifying weighted-graph | |
CN103903013A (en) | Optimization algorithm of unmarked flat object recognition | |
CN105931253A (en) | Image segmentation method combined with semi-supervised learning | |
CN107527054B (en) | Automatic foreground extraction method based on multi-view fusion | |
CN104615996B (en) | A kind of various visual angles two-dimension human face automatic positioning method for characteristic point | |
CN102289812A (en) | Object segmentation method based on priori shape and CV (Computer Vision) model | |
CN106326924A (en) | Object tracking method and object tracking system based on local classification | |
Adoram et al. | IRUS: image retrieval using shape | |
CN106056627A (en) | Robustness object tracking method based on local identification sparse representation | |
CN103093211B (en) | Based on the human body motion tracking method of deep nuclear information image feature | |
CN107330434A (en) | Electrical symbol recognition methods in a kind of circuit diagram based on PHOG features | |
CN104253994A (en) | Night monitored video real-time enhancement method based on sparse code fusion | |
CN110047085A (en) | A kind of accurate restorative procedure in lung film coalescence knuckle areas for lung CT carrying out image threshold segmentation result | |
CN103093241B (en) | Based on the remote sensing image nonuniformity cloud layer method of discrimination of homogeneity process | |
CN105069766A (en) | Inscription restoration method based on contour feature description of Chinese character image | |
CN110188830B (en) | SAR image change detection method based on multi-core graph cut | |
CN103871055B (en) | A kind of well-marked target detection method based on dynamic anisotropy receptive field | |
Zhang et al. | An improved adaptive level set method for image segmentation | |
Ma et al. | A real time detection method of track fasteners missing of railway based on machine vision | |
Wang et al. | Cell detection and tracking for micromanipulation vision system of cell-operation robot |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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 | ||
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: 20190712 Termination date: 20201104 |