CN106920244B - A kind of method of the neighbouring background dot of detection image edges of regions - Google Patents

A kind of method of the neighbouring background dot of detection image edges of regions Download PDF

Info

Publication number
CN106920244B
CN106920244B CN201710025201.0A CN201710025201A CN106920244B CN 106920244 B CN106920244 B CN 106920244B CN 201710025201 A CN201710025201 A CN 201710025201A CN 106920244 B CN106920244 B CN 106920244B
Authority
CN
China
Prior art keywords
point
class
subscript
image
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.)
Active
Application number
CN201710025201.0A
Other languages
Chinese (zh)
Other versions
CN106920244A (en
Inventor
高理文
林小桦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangzhou University of Chinese Medicine
Original Assignee
Guangzhou University of Chinese Medicine
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangzhou University of Chinese Medicine filed Critical Guangzhou University of Chinese Medicine
Priority to CN201710025201.0A priority Critical patent/CN106920244B/en
Publication of CN106920244A publication Critical patent/CN106920244A/en
Application granted granted Critical
Publication of CN106920244B publication Critical patent/CN106920244B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10004Still image; Photographic image

Abstract

The invention discloses a kind of method of background dot near detection image edges of regions, steps are as follows: firstly, sampling is chosen using region approximation marginal point as the point on the circumference in the center of circle;Then, the point on circumference is arranged, is stored in list.Then, different start-stop address marks is converted repeatedly, the point in list is divided into two class of foreground point and background dot, calculates sorted comprehensive inter-class variance every time and is corrected.In turn, the corresponding classification schemes of comprehensive correction to variances value the maximum are chosen, with its start-stop address mark, calculate the position for determining required background dot.

Description

A kind of method of the neighbouring background dot of detection image edges of regions
Technical field
The present invention relates to image recognition research field, in particular to the side of background dot near a kind of detection image edges of regions Method.
Background technique
For a long time, image segmentation is the research hotspot in Image Engineering.Under normal circumstances, threshold method, edge detection, area The methods of domain growth is simple and effective.But for the more fuzzy complex background image in many boundaries, watershed side Method, movable contour model etc. are more applicable.And at this point, if the background dot of several adjacent edges can be obtained first, it will be significantly advantageous In the raising for promoting segmentation accuracy.
It is related to detection a little, it has to lift the feature point detecting methods such as SIFT, SURF.But they are all to extract image In characteristic point, in region or region outside;It is chiefly used in the registration of image.
There are also certain methods to extract the characteristic point on edge.Such as Antonio Louro in 2012 is proposed in bianry image Vital point on middle detection edge, to describe region shape information entrained by edge.For another example U.A.A. in 2016 Niroshika etc. proposes a kind of isotropic method, and the angle point on edge is detected in gray level image.
A kind of detection target significant point proposed there are also noble soldiers in 2016 that merits attention simultaneously generates initial profile then The method for carrying out image segmentation.Significant boundary point is obtained by the Harris operator that color is promoted first, secondly proposes core Notable figure obtains target seed point from significant boundary point, and then the significant boundary point of target is determined by these seed points, last mesh Significant boundary point is marked as the seed point of convex closure and generates initial profile and in this, as LRAC model (localized areas type castor Wide model) initial profile to image segmentation.But many hypothesis are used in this method, it is assumed for example that target pixel points and background dot Between color difference it is very big.Obviously this is exactly that many complex background image institutes are unappeasable.
Summary of the invention
The present invention in order to overcome at least one of the drawbacks of the prior art described above (deficiency), provides a kind of detection image region The method of adjacent edges background dot.Background dot near the edges of regions obtained using this method can be directly used in the complicated back of promotion The accuracy of scape image segmentation.For example, can be used as the context marker of label dividing ridge method;To effectively prevent watershed to build Outside target area.
There are two preconditions for this method:
(1) an approximate edges of regions point known to.
(2) it can determine that the approximate region marginal point nearby belongs in target area in some region.
In order to solve the above technical problems, technical scheme is as follows:
A kind of method of the neighbouring background dot of detection image edges of regions, comprising the following steps:
S1, note present image are currentImage, input point p and prospect binary map foregroundMask;
S2, centered on point p, on the circumference that radius is radius, by clockwise or counterclockwise, equably choose CirclePointsNum point successively stores the abscissa of each point and the binary group of ordinate composition in circlePoints;
The pixel value arrangement duplication of S3, circumferential point;Specifically: each of sequence detection circlePoints is put preceding Value in scape binary map foregroundMask, from circlePoints, in prospect binary map foregroundMask It is denoted as the point of " 1 ", is extracted from circumference in order, their red, green, blue pixels in image currentImage are successively replicated It is worth pixelList, while records corresponding position of their transverse and longitudinal coordinate into pixelIndexInCirclePoints;And Afterwards, from circlePoints, the point for being denoted as " 0 " in prospect binary map foregroundMask, in order from circumference It extracts, the last one point on followed by pixelList, successively replicates red, green, blue pixel of each point in currentImage It is worth in pixelList, while records corresponding position of their transverse and longitudinal coordinate in pixelIndexInCirclePoints; It records in pixelList, is " 1 " and the subscript of that maximum point of subscript in foregroundMask LastInpointIndex;
S4, initialization maximum between-cluster variance maxD=- 1;
S5, initial background class starting subscript i are equal to LastInpointIndex+1;
S6, judge whether background classes starting subscript i is less than or equal to circlePointsNum;If so, turning to step S7;It is no Then, step S17 is turned to;
S7, initial background class terminate subscript j and are equal to background classes starting subscript i;
S8, judge that background classes terminate whether subscript j is less than or equal to circlePointsNum;If so, turning to step S9;It is no Then, step S16 is turned to;
S9, seek component inter-class variance: in pixelList, point of the subscript from i to j regards one kind as;Remaining is another Class asks inter-class variance dr, dg, db with regard to three components of red pixel value, green pixel values, blue pixel value of 2 class points respectively; Wherein inter-class variance formula are as follows: d=ω 1 (μ 1- μ) ^2+ ω 2 (μ 2- μ) ^2;
Wherein, 1 ω, ω 2 are respectively the probability of the point appearance of the first kind and the second class, ω 1+ ω 2=1;μ indicates all the points The pixel mean value of a certain color component;μ 1, μ 2 are respectively the pixel mean value of a certain color component of point of the first kind and the second class;
S10, comprehensive inter-class variance D=dr+dg+db is enabled;
S11, correction factor modifyCoefficient is sought, specifically:
Step11.1: the mean value of i and j and rounding are calculated, midOfij is denoted as;
Step11.2: the minimum in midOfij-LastInpointIndex and circlePointsNum-midOfij is sought Person is denoted as testX;
Step11.3: it calculates (circlePointsNum-LastInpointIndex)/2, and is rounded, be denoted as testR;
Step11.4:modifyCoefficient=arctan (testX/testR × 20)/(pi/2);Arctan is anyway Cut function;
S12, comprehensive correction to variances value D ', D '=D^modifyCoefficient is sought based on following formula;
S13, judge whether current D ' is greater than maxD, if so, turning to step S14;Otherwise, step S15 is turned to;
S14, optimal value: maxD=D ' is updated;bestStartIndex=i;bestStopIndex=j;Wherein BestStartIndex indicates that the background dot class of optimal classification scheme originates subscript;BestStopIndex indicates optimal classification side The background dot class of case terminates subscript;
S15, j=j+1 is updated;Turn to step S8;
S16, i=i+1 is updated;Turn to step S6;
The coordinate of background dot o near point p required by S17, determination: to bestStartIndex and bestStopIndex It averages and is rounded, as subscript, inquiry obtains point o in image in pixelIndexInCirclePoints Transverse and longitudinal coordinate in currentImage.
Preferably, in step S1, point p is an approximate marginal point of target area in present image, the approximation edge Point refers to that the shortest distance from target area edge is less than the pixel of present image short side M%;Prospect binary map ForegroundMask is one and is included in the sub- foreground area of target area and is located near point p, the prospect two-value Figure foregroundMask is located at the N% for referring to that the shortest distance between the two is less than present image short side near point p;
The value range of M be (0,3];
The value range of N be (0,3];
Radius is greater than the M% and N% of present image currentImage short side, and is less than present image currentImage The 10% of short side;
The value range of circlePointsNum is radius [2,12] times and is rounded.
Generally speaking, the present invention is directed to the complex background image for being difficult to divide, and proposes that a kind of detection image edges of regions is attached The method of nearly background dot, the background dot near edges of regions mentioned here refer to that be less than image at a distance from edges of regions short The background dot of the M% on side.Background dot near edges of regions can be used to inhibit the mistake expansion that region is outside in image segmentation process. Certainly it is also not excluded for can be used for certain features of detection zone adjacent edges background.
Compared with prior art, the beneficial effect of technical solution of the present invention is:
The background dot near approximate region marginal point can be relatively accurately detected, to be conducive to complex background image Accurate segmentation or its adjacent edges background characteristics detection.Specifically, this method is directly based upon color image, ash is avoided Information caused by degreeization is lost;Two classes are carried out using the series of points on maximum variance between clusters pairing approximation marginal point neighborhood circumference Classification can effectively improve its noise immunity compared to the gradient etc. for calculating a single point.
Detailed description of the invention
Fig. 1 is the flow chart of the method for the present invention.
Fig. 2 (a) is bignose rhinacanthus branchlet and leaf leaf image
Fig. 2 (b) is the position view of p point in bignose rhinacanthus branchlet and leaf leaf image.
Fig. 2 (c) is the schematic diagram of foregroundMask.
The schematic diagram of the background dot o of Fig. 2 (d) leaf adjacent edges.
Specific embodiment
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent;In order to better illustrate this embodiment, attached Scheme certain components to have omission, zoom in or out, does not represent the size of actual product;
To those skilled in the art, it is to be understood that certain known features and its explanation, which may be omitted, in attached drawing 's.The following further describes the technical solution of the present invention with reference to the accompanying drawings and examples.
A kind of method of the neighbouring background dot of detection image edges of regions, includes the following steps, such as Fig. 1: firstly, sampling is chosen Using region approximation marginal point as the point on the circumference in the center of circle;Then, the point on circumference is arranged, is stored in list.Then, Different start-stop address marks is converted, the point in list is divided into two class of foreground point and background dot, its inter-class variance is calculated and repairs Just.In turn, the corresponding classification schemes of amendment inter-class variance the maximum are chosen to calculate required by determining with its start-stop address mark The position of background dot.
Its detailed process are as follows:
Step1: Fig. 2 (a) show current currently processed complex background leaf image, is denoted as currentImage;As shown in Fig. 2 (b), a note in currentImage in four leaf marginal points of user's continued labelling For point p;As shown in Figure 2 (c), the point of four continued labellings is sequentially connected the region that composed quadrangle is included, as The known sub- foreground area for being contained in target area near point p, is denoted as prospect binary map foregroundMask;
The acquisition of point p can be demarcated by user.The solution of prospect binary map foregroundMask is based on different items Part has and different seeks method.For example, if user demarcate three with up contour point, and ensure all marginal points be sequentially connected gained Region in target area, then, region obtained by connection above-mentioned is exactly foregroundMask.In another example using harshness Threshold value, the every bit near point P is differentiated, segmentation obtain foregroundMask.
Step2: the point on circumference is chosen in sampling.Centered on point p, radius(as 25) for radius circumference on, by suitable Hour hands or counter clockwise direction equably choose such as circlePointsNum=radius × 9 circlePointsNum() a point, The abscissa of each point and the binary group (back abbreviation transverse and longitudinal coordinate) of ordinate composition are successively stored in circlePoints.
Step3: the pixel value of circumferential point arranges duplication.Each of sequence detection circlePoints is put in prospect two Value in value figure foregroundMask.From circlePoints, in prospect binary map foregroundMask be " 1 " Part point, the sequence extracted from circumference by it successively replicates their red, green, blue pixels in currentImage It is worth in pixelList, while records same position of their transverse and longitudinal coordinate into pixelIndexInCirclePoints (i.e. The pixel value of certain point copies to which of pixelList, its transverse and longitudinal coordinate just copies to Which position of pixelIndexInCirclePoints).Then, from circlePoints, in prospect binary map It is the part point of " 0 " in foregroundMask, the sequence extracted from circumference by it, on followed by pixelList most The latter point successively replicates their red, green, blue pixel values in currentImage in pixelList, while recording them Same position of the transverse and longitudinal coordinate in pixelIndexInCirclePoints.It records in pixelList, It is the subscript LastInpointIndex of " 1 " and that maximum point of subscript in foregroundMask.
Step4: initialization maximum between-cluster variance maxD=- 1.
Step5: initial background class originates subscript i and is equal to LastInpointIndex+1.
Step6: judge whether background classes starting subscript i is less than or equal to circlePointsNum.If so, turning to step step7;Otherwise, two are completed to recirculate, turns to step17.
Step7: initial background class terminates subscript j and is equal to background classes starting subscript i.
Step8: judge that background classes terminate whether subscript j is less than or equal to circlePointsNum.If so, turning to step step9;Otherwise, step16 is turned to.
Step9: seek component inter-class variance: in pixelList, point of the subscript from i to j regards one kind as;Remaining is another One kind asks inter-class variance dr, dg, db respectively to red pixel value, green pixel values, blue pixel value.Inter-class variance formula Are as follows: d=ω 1 (μ 1- μ) ^2+ ω 2 (μ 2- μ) ^2.
Wherein, 1 ω, ω 2 are respectively the probability of the point appearance of the first kind and the second class, ω 1+ ω 2=1;μ indicates all the points The pixel mean value of a certain color component;μ 1, μ 2 are respectively the pixel mean value of a certain color component of point of the first kind and the second class.It is public Formula d=ω 1 (μ 1- μ) ^2+ ω 2 (μ 2- μ) ^2 is respectively to red, green, blue, using three times.Such as, for the first time, all foreground points Red pixel value is the first kind, and the red pixel value of all background dots is the second class, then seeks its inter-class variance using formula.
Step10: comprehensive inter-class variance D=dr+dg+db is enabled.
Step11: seeking correction factor modifyCoefficient, specifically:
Step11.1: the mean value of i and j and rounding are calculated, midOfij is denoted as.
Step12.11.2: it asks in midOfij-LastInpointIndex and circlePointsNum-midOfij most Small person, is denoted as testX.
Step11.3: it calculates (circlePointsNum-LastInpointIndex)/2, and is rounded, be denoted as testR.
Step11.4:modifyCoefficient=arctan (testX/testR × 20)/(pi/2);Arctan is anyway Cut function.
The point for being wherein designated as midOfij down is quasi-definite leaf exterior point o;Leaf area includes prospect marked region, i.e., The region that point in foregroundMask for " 1 " forms,
Illustrate: under to be designated as the point of midOfij be quasi-definite leaf exterior point o.Leaf area includes prospect marked region (region that the point in foregroundMask for " 1 " forms), the former range is bigger than the latter.Singly for this factor, A possibility that two boundary points of the point o apart from prospect marked region are remoter, it correctly falls on leaf exterior domain is bigger.When point o is non- When very close to prospect marked region, even if it is still in the outside of prospect marked region, still, it is mistakenly fallen into blade The risk in portion is also very high.So generally speaking, the effect of correction factor modifyCoefficient is encouraged far from prospect mark Remember the situation in region, inhibits the situation too close to prospect marked region.
Step12: comprehensive correction to variances value D ', D '=D^modifyCoefficient is asked based on following formula.
S13, judge whether current D ' is greater than maxD, if so, turning to step S14;Otherwise, step S15 is turned to;
S14, optimal value: maxD=D ' is updated;bestStartIndex=i;bestStopIndex=j;Wherein BestStartIndex indicates that the background dot class of optimal classification scheme originates subscript;BestStopIndex indicates optimal classification side The background dot class of case terminates subscript;
Step15:j=j+1.Turn to step8.
Step16:i=i+1.Turn to step6.
Step17: the coordinate of the background dot o near required point p is determined, as shown in Figure 2: to bestStartIndex It averages and is rounded with bestStopIndex, as subscript, inquire and obtained a little in pixelIndexInCirclePoints Transverse and longitudinal coordinate of the o in currentImage.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this Made any modifications, equivalent replacements, and improvements etc., should be included in the claims in the present invention within the spirit and principle of invention Protection scope within.

Claims (2)

1. a kind of method of background dot near detection image edges of regions, which comprises the following steps:
S1, note present image are currentImage, input point p and prospect binary map foregroundMask;Point p is currently to scheme The approximate marginal point of one of target area as in, the approximation marginal point refer to that the shortest distance from target area edge is less than and work as The pixel of preceding image short side M%, the value range of M be (0,3];
S2, centered on point p, on the circumference that radius is radius, by clockwise or counterclockwise, equably choose CirclePointsNum point successively stores the abscissa of each point and the binary group of ordinate composition in circlePoints;
The pixel value arrangement duplication of S3, circumferential point;Specifically: each of sequence detection circlePoints is put in prospect two Value in value figure foregroundMask, from circlePoints, being denoted as in prospect binary map foregroundMask The point of " 1 ", is extracted from circumference in order, is successively replicated their red, green, blue pixel values in image currentImage and is arrived PixelList, while recording corresponding position of their transverse and longitudinal coordinate into pixelIndexInCirclePoints;Then, From circlePoints, the point for being denoted as " 0 " in prospect binary map foregroundMask, taken out from circumference in order It takes, the last one point on followed by pixelList, successively replicates red, green, blue pixel value of each point in currentImage Into pixelList, while recording corresponding position of their transverse and longitudinal coordinate in pixelIndexInCirclePoints;Note It records in pixelList, is " 1 " and the subscript of that maximum point of subscript in foregroundMask LastInpointIndex;
S4, initialization maximum between-cluster variance maxD=-1;
S5, initial background class starting subscript i are equal to LastInpointIndex+1;
S6, judge whether background classes starting subscript i is less than or equal to circlePointsNum;If so, turning to step S7;Otherwise, turn To step S17;
S7, initial background class terminate subscript j and are equal to background classes starting subscript i;
S8, judge that background classes terminate whether subscript j is less than or equal to circlePointsNum;If so, turning to step S9;Otherwise, turn To step S16;
S9, seek component inter-class variance: in pixelList, point of the subscript from i to j regards one kind as;It is remaining to be another kind of, point Inter-class variance dr, dg, db are not asked with regard to three components of red pixel value, green pixel values, blue pixel value of 2 class points;Wherein class Between formula of variance are as follows: d=ω 1 (μ 1- μ) ^2+ ω 2 (μ 2- μ) ^2;
Wherein, 1 ω, ω 2 are respectively the probability of the point appearance of the first kind and the second class, ω 1+ ω 2=1;μ indicates that all the points are a certain The pixel mean value of color component;μ 1, μ 2 are respectively the pixel mean value of a certain color component of point of the first kind and the second class;
S10, comprehensive inter-class variance D=dr+dg+db is enabled;
S11, correction factor modifyCoefficient is sought, specifically:
Step11.1: the mean value of i and j and rounding are calculated, midOfij is denoted as;
Step11.2: seeking the reckling in midOfij-LastInpointIndex and circlePointsNum-midOfij, note For testX;
Step11.3: it calculates (circlePointsNum-LastInpointIndex)/2, and is rounded, be denoted as testR;
Step11.4:modifyCoefficient=arctan (testX/testR × 20)/(pi/2);Arctan is arc tangent Function;
S12, comprehensive correction to variances value D ', D '=D^modifyCoefficient is sought based on following formula;
S13, judge whether current D ' is greater than maxD, if so, turning to step S14;Otherwise, step S15 is turned to;
S14, optimal value: maxD=D ' is updated;BestStartIndex=i;BestStopIndex=j;Wherein BestStartIndex indicates that the background dot class of optimal classification scheme originates subscript;BestStopIndex indicates optimal classification side The background dot class of case terminates subscript;
S15, j=j+1 is updated;Turn to step S8;
S16, i=i+1 is updated;Turn to step S6;
The coordinate of background dot o near point p required by S17, determination: equal is asked to bestStartIndex and bestStopIndex It is worth and is rounded, as subscript, inquiry obtains point o in image currentImage in pixelIndexInCirclePoints In transverse and longitudinal coordinate.
2. the method for background dot near detection image edges of regions according to claim 1, which is characterized in that step S1 In, point p is an approximate marginal point of target area in present image, and the approximation marginal point refers to from target area edge The shortest distance is less than the pixel of present image short side M%;Prospect binary map foregroundMask is one and is included in target The sub- foreground area in region and be located at point p near, the prospect binary map foregroundMask, which is located near point p, is Refer to N% of the shortest distance between the two less than present image short side;
The value range of M be (0,3];
The value range of N be (0,3];
Radius is greater than the M% and N% of present image currentImage short side, and it is short to be less than present image currentImage The 10% of side;
The value range of circlePointsNum is radius [2,12] times and is rounded.
CN201710025201.0A 2017-01-13 2017-01-13 A kind of method of the neighbouring background dot of detection image edges of regions Active CN106920244B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710025201.0A CN106920244B (en) 2017-01-13 2017-01-13 A kind of method of the neighbouring background dot of detection image edges of regions

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710025201.0A CN106920244B (en) 2017-01-13 2017-01-13 A kind of method of the neighbouring background dot of detection image edges of regions

Publications (2)

Publication Number Publication Date
CN106920244A CN106920244A (en) 2017-07-04
CN106920244B true CN106920244B (en) 2019-08-02

Family

ID=59453490

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710025201.0A Active CN106920244B (en) 2017-01-13 2017-01-13 A kind of method of the neighbouring background dot of detection image edges of regions

Country Status (1)

Country Link
CN (1) CN106920244B (en)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102779339A (en) * 2011-12-31 2012-11-14 北京京东方光电科技有限公司 Image processing method and system
CN103119609A (en) * 2012-09-27 2013-05-22 华为技术有限公司 Method and device for determining video foreground main image area
CN103198470A (en) * 2013-02-26 2013-07-10 清华大学 Image cutting method and image cutting system
CN103530893A (en) * 2013-10-25 2014-01-22 南京大学 Foreground detection method in camera shake scene based on background subtraction and motion information
CN103618846A (en) * 2013-11-22 2014-03-05 上海安奎拉信息技术有限公司 Background removing method for restricting influence of sudden changes of light in video analysis
CN103824297A (en) * 2014-03-07 2014-05-28 电子科技大学 Multithreading-based method for quickly updating background and foreground in complex high dynamic environment
CN104778721A (en) * 2015-05-08 2015-07-15 哈尔滨工业大学 Distance measuring method of significant target in binocular image

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102779339A (en) * 2011-12-31 2012-11-14 北京京东方光电科技有限公司 Image processing method and system
CN103119609A (en) * 2012-09-27 2013-05-22 华为技术有限公司 Method and device for determining video foreground main image area
CN103198470A (en) * 2013-02-26 2013-07-10 清华大学 Image cutting method and image cutting system
CN103530893A (en) * 2013-10-25 2014-01-22 南京大学 Foreground detection method in camera shake scene based on background subtraction and motion information
CN103618846A (en) * 2013-11-22 2014-03-05 上海安奎拉信息技术有限公司 Background removing method for restricting influence of sudden changes of light in video analysis
CN103824297A (en) * 2014-03-07 2014-05-28 电子科技大学 Multithreading-based method for quickly updating background and foreground in complex high dynamic environment
CN104778721A (en) * 2015-05-08 2015-07-15 哈尔滨工业大学 Distance measuring method of significant target in binocular image

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
三维电视中虚拟视点合成的技术研究;王路;《中国优秀硕士学位论文全文数据库-信息科技辑》;20130215(第2期);第I138-1476页
融合视觉模型和最大类间方差的阈值分割算法;邹小林 等;《计算机应用》;20130301;第33卷(第3期);第670-673+837页

Also Published As

Publication number Publication date
CN106920244A (en) 2017-07-04

Similar Documents

Publication Publication Date Title
US9235762B2 (en) Iris data extraction
Wu et al. A novel color image segmentation method and its application to white blood cell image analysis
CN105117726B (en) License plate locating method based on multiple features zone-accumulation
CN103295018B (en) A kind of branches and leaves block fruit precise recognition method
CN104182756B (en) Method for detecting barriers in front of vehicles on basis of monocular vision
CN104392240A (en) Parasite egg identification method based on multi-feature fusion
CN106875394A (en) Separation and group technology suitable for the phase contrast microscope cell main information adhesion situation without inter-reference information
CN102999916A (en) Edge extraction method of color image
CN106920244B (en) A kind of method of the neighbouring background dot of detection image edges of regions
CN105654085A (en) Image technology-based bullet hole recognition method
CN104778442A (en) Automatic cutting and counting method for fluorescent microscopic images of retinal cells
CN103680145B (en) A kind of people's car automatic identifying method based on local image characteristics
CN102214355A (en) Cutout method of clothing display material
CN102663700A (en) Segmentation method for adhering grain binary image
CN106910197B (en) A kind of dividing method of the complex background leaf image in single goal region
CN106503704A (en) Circular traffic sign localization method in a kind of natural scene
CN105701818A (en) Multi-target image segmentation C-V method based on area division and gradient guiding
CN107767383A (en) A kind of Road image segmentation method based on super-pixel
CN107578035A (en) Human body contour outline extracting method based on super-pixel polychrome color space
JP2002083301A (en) Traffic monitoring device
CN108537217B (en) Character coding mark-based identification and positioning method
CN105184817A (en) Moving object detection method by overcoming static foreground
CN104298966A (en) License plate positioning method
CN107330881A (en) A kind of method for recognizing wheel hub valve port
CN104616007B (en) A kind of vehicle identification method based on conspicuousness detection and color histogram graph model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant