CN109598726A - A kind of adapting to image target area dividing method based on SLIC - Google Patents

A kind of adapting to image target area dividing method based on SLIC Download PDF

Info

Publication number
CN109598726A
CN109598726A CN201811258689.2A CN201811258689A CN109598726A CN 109598726 A CN109598726 A CN 109598726A CN 201811258689 A CN201811258689 A CN 201811258689A CN 109598726 A CN109598726 A CN 109598726A
Authority
CN
China
Prior art keywords
pixel
super
point
image
segmentation
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.)
Pending
Application number
CN201811258689.2A
Other languages
Chinese (zh)
Inventor
于天河
王成栋
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.)
Harbin University of Science and Technology
Original Assignee
Harbin University of Science and Technology
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 Harbin University of Science and Technology filed Critical Harbin University of Science and Technology
Priority to CN201811258689.2A priority Critical patent/CN109598726A/en
Publication of CN109598726A publication Critical patent/CN109598726A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

A kind of adapting to image target area dividing method based on SLIC, is related to super-pixel segmentation technology.Purpose be in order to solve the problem of traditional superpixel segmentation method to comprising multiple targets subject image carry out super-pixel segmentation when need to be manually entered parameter cause divide low efficiency.The invention firstly uses SLIC to carry out super-pixel pre-segmentation processing to image, and then point establishes super-pixel unit centered on the good super-pixel point of pre-segmentation, and super-pixel linear module includes gray scale, position, cryptographic Hash.Use this linear module as super-pixel point parameter, the clustering algorithm by auto-adaptive parameter based on distance will finally be partitioned into zonule again and merge, and then super-pixel is clustered into determining main body and apparent main body partitioning boundary.This method does not need user and is configured input, determines the number for needing to divide super-pixel type by the method for calculating image complexity.Suitable for fields such as target identification, pattern-recognition, artificial intelligence.

Description

A kind of adapting to image target area dividing method based on SLIC
Technical field
The present invention relates to image processing techniques more particularly to super-pixel segmentation technologies.
Background technique
Recent years, people are increasingly hot to the research of super-pixel, especially at many aspects of machine vision, all start One agitation to super-pixel research, image segmentation have a very wide range of applications in actual processing, know in target Not, the various fields such as pattern-recognition, artificial intelligence have obtained universal application.But as picture size is increasing, directly exist The computational efficiency handled in pixel granularity level to image is lower, this requires reducing pixel quantity, expands pixel institute's generation The meaning of table, so just needing to replace on the basis of original pixel using super-pixel, super-pixel segmentation is exactly pixel cluster At the process of super-pixel.It saves the boundary information of image by the polymerization of pixel, reduces the complexity of subsequent image processing Degree.Super-pixel segmentation algorithm accelerates the speed of image procossing, and remains boundary information well.It can accelerate to handle in this way Speed, on the basis of guaranteeing processing accuracy also to guarantee processing speed.Image procossing on the basis of it becomes very Good research direction, has in terms of image segmentation, target detection, feature extraction, target following, human posture at present Good development.But for the subject image comprising multiple targets, need that parameter is manually entered in super-pixel segmentation, Setting segmentation number, limits segmentation efficiency.
Summary of the invention
The purpose of the present invention is to solve traditional superpixel segmentation methods to the subject image comprising multiple targets When carrying out super-pixel segmentation, due to needing to be manually entered parameter, setting segmentation number leads to the problem of dividing low efficiency, provides One kind being based on the adapting to image target area dividing method of SLIC (simple linear iteration cluster).
Dividing method that the adapting to image target area based on SLIC merges the following steps are included:
Step 1: reading original image, down-sampling processing is carried out to the original image, down-sampling step-length is S pixel list Position, i.e., all pixels point by original image in the sliding window of S*S merges into a super-pixel point, in the super-pixel point All pixels point in, the pixel of average gray value that gray value is all pixels point in the super-pixel point is surpassed into picture as this The representative of vegetarian refreshments；
Step 2: being initialized to the image seed point number after down-sampling, according to the seed point after initialization Number, seed point is evenly distributed in the various pieces of image；
Step 3: the shade of gray for calculating each pixel in the range of the n*n centered on seed point (generally takes n =2) the smallest pixel of shade of gray, is then found out, using the position of the pixel as the final position of seed point；
Step 4: carrying out classification marker to the pixel around the seed point in step 4, specifically include:
Step 4 one scans for the pixel around seed point, and moving step length used in search process is Step =sqrt (N/K)+q, wherein N is the quantity of pixel in original image, and K is the quantity of super-pixel point after segmentation, and q is positive whole Number；
Step 4 two, calculating each pixel in search range, at a distance from seed point, which includes Gray homogeneity d1、 Space length d2With induction Hash distance d3,
Wherein m, n are the pixel number in corresponding super-pixel block respectively, and k is the gray value of pixel；
Wherein x, y are the abscissa and ordinate of corresponding pixel points respectively；
Wherein, X indicates seed point and difference of the neighbouring super pixels point in gray scale on its up and down four direction；
Step 4 three, the total variances value D for calculating each pixel and seed point in search range:
D=μ1d1+μ2d2+μ3d3
Wherein μ1、μ2And μ3Respectively represent the weight coefficient of each distance, and μ1、μ2、μ3Meet following relationship:
μ1=μ2=5 μ3=log2K
Total variances value is compared step 4 four with according to preset threshold value, and it is poly- to determine which pixel belongs to this Class pixel dot center；
Step 5: multiplicating step 3 and step 4 so far complete image until cluster result does not change Segmentation.
Further, the method also includes:
Step 6: enhancing the connectivity of image after segmentation.
Further, the S=2 in step 1.
Further, the q=2 in step 4 one.
Further, the number of repetition in step 5 is 10.
The above method carries out super-pixel pre-segmentation processing to image first with SLIC, and original image is divided into size phase Seemingly, the super-pixel point of regular shape, then point establishes super-pixel unit, super-pixel measurement centered on the super-pixel point divided Unit includes gray scale, position, cryptographic Hash.It uses this linear module as super-pixel point parameter, distance is based on by auto-adaptive parameter Clustering algorithm, will finally be partitioned into zonule again and merge, and then super-pixel is clustered into determining main body and apparent Main body partitioning boundary.This method does not need user and is configured input, determines needs by the method for calculating image complexity Divide the number of super-pixel type.The experimental results showed that the algorithm can be solved effectively in the segmentation of subject image containing multiple target The problem of manually carrying out input parameter is needed, segmentation number can be adaptively determined under the premise of guaranteeing segmentation effect, is mentioned Time needed for height segmentation efficiency, shortening image procossing.
Detailed description of the invention
Fig. 1 is the flow chart of the adapting to image target area dividing method of the present invention based on SLIC；
(a) in Fig. 2 is by step 1 treated picture, (b) (c) to be by step 2 treated picture By step 7 treated picture；
(a) in Fig. 3 is original image, (b) for by step 6 treated picture；
(a) in Fig. 4 is original image, (b) for by step 6 treated picture.
Specific embodiment
Specific embodiment 1: as shown in Figures 1 and 2, the adapting to image mesh described in present embodiment based on SLIC Mark region segmentation method comprises the steps of:
Step 1: reading original image, down-sampling processing is carried out to the original image, down-sampling step-length is S pixel list Position, i.e., all pixels point by original image in the sliding window of S*S merges into a super-pixel point, this pixel Parameter value is exactly the mean value that all pixels point corresponds to parameter in window.In all pixels point in the super-pixel point, by gray scale Value is representative of the pixel of the average gray value of all pixels point in the super-pixel point as the super-pixel point, i.e., in window The average gray value of pixel is that the pixel of gray value goes the representative of all pixels point as this window area.Such as original image As 100*100, total 10000 pixels need to handle, and original image reforms into 50*50 after the method, amount to 2500 The time of image procossing is greatly reduced in a pixel.For balanced division effect and image processing speed, the preferred value of S is 2；
Pk=∑i∈win(k)Ii/S2
Wherein, I is pixel original attribute value, and i is the position number of each pixel in window, PkFor pixel category each in window Property average value, in the method, belonging to originally property refer to gray scale.
Step 2: (cluster centre) is initialized to the image seed point number after down-sampling, after initialization Seed point number, seed point is evenly distributed in the various pieces of image；Assuming that including the picture of N number of pixel, in advance Setting is divided into K super-pixel point (K value is generally 500-1000), and each super-pixel point size is identical, then a super-pixel point Magnitude range be exactly N/K, similarly can approximately calculate each super-pixel point moving step length be S=sqrt (N/K).This One step is that the pixel in picture is presorted, and increases the connection between neighbor pixel.
Step 3: initializing cluster seed point number again, seed point location is upset, theme is exactly in seed point to be selected Region searches out the position being most suitable for as seed point.Specific method is to calculate in the range of the n*n centered on seed point The shade of gray (generally taking n=2) of each pixel, then finds out the smallest pixel of shade of gray, by the position of the pixel The final position as seed point is set (using the coordinate of the pixel as the position attribution of this super-pixel point, in distance below It is used when measurement)；Because the change of gradient of boundary pixel is usually bigger, seed can be effectively avoided in this way Point is fallen on the edge of super-pixel segmentation, is influenced the cluster of subsequent similar super-pixel, is caused segmentation effect poor.
Step 4: carrying out super-pixel metric calculation to the pixel around the seed point in step 4, classification polymerization is carried out, Which cluster pixel dot center namely belonged to, threshold value setting is different, and the sparse degree of seed point super-pixel classification is also different, tool Body includes:
Step 4 one scans for the pixel around seed point, and moving step length used in search process is Step =sqrt (N/K)+q, wherein N is the quantity of pixel in original image, and K is the quantity of super-pixel point after segmentation, and q is positive whole Number, the preferred value of q are 2, and adding q is that super-pixel is too small in order to prevent, the situation for causing the super-pixel block being partitioned into too intensive.One As the search range of sub-pixel point be exactly 2 times of step-length range, but it is desirable to the super-pixel size generated is single times of step-length；
Step 4 two, using in step 1 representative calculate search range in each pixel at a distance from seed point, should Distance includes Gray homogeneity d1, space length d2With induction Hash distance d3,
Wherein m, n are the pixel number in corresponding super-pixel block respectively, and k is the gray value of pixel；
Wherein x, y are the abscissa and ordinate of corresponding pixel points respectively；
Wherein, X indicates seed point and difference of the neighbouring super pixels point in gray scale on its up and down four direction, difference Not beyond being denoted as 0, it is denoted as 1 beyond threshold value, then generates one four binary codes.Due to being related to up and down totally four sides To, therefore the value of i is 0,1,2,3；
In the step, gray count method is the average value for calculating the gray scale of each pixel in super-pixel block, and position is meter Calculate the geometric average point coordinate of each super-pixel, cryptographic Hash be calculate super-pixel block the point of super-pixel up and down and its own in ash Difference on degree generates one four binary codes by certain threshold value.Difference between traditional calculating pixel is neglected The pixel and the difference of its own around pixel have been omited, has only focused on the attribute difference between pixel itself, has ignored pixel The difference index of point and its surrounding pixel point.Be added after this index can be more obvious distinguish discrepant pixel simultaneously Classify to it.Preferably the super-pixel point with similar state can be condensed together.And be experimentally confirmed, this Method can make segmentation result have this better sensitivity for boundary, more sensitive to boundary.
Step 4 three, the total variances value D for calculating each pixel and seed point in search range:
D=μ1d1+μ2d2+μ3d3
Wherein μ1、μ2And μ3Respectively represent the weight coefficient of each distance, and μ1、μ2、μ3Meet following relationship:
μ1=μ2=5 μ3=log2K
K can be understood as pre-segmentation super-pixel point type, and type is much more complicated with regard to representative image, and image is more complicated with regard to generation Table detail section to be processed is more, and the difference criteria of super-pixel point just needs to be amplified in image, therefore weight coefficient and figure As complexity positive correlation.Impact factor around super-pixel point cannot be excessive, it is excessive be easy for protruding it is some tiny The influence on boundary impacts last segmentation object, and most general objective is by picture segmentation into object to be measured class.
Total variances value is compared step 4 four with according to preset threshold value, and it is poly- to determine which pixel belongs to this Class pixel dot center；
Step 5: multiplicating step 3 and step 4 so far complete image until cluster result does not change Segmentation.As shown in Figures 3 and 4.Theoretically constantly iteration is updated through the above steps until error convergence, i.e. iteration update directly Until cluster result does not change.By largely practicing discovery, most pictures can obtain more by iteration 10 times Ideal effect, therefore the number of iterations generally takes 10.
Step 6: enhancing the connectivity of image after segmentation.Following flaw is likely to occur by above-mentioned iteration optimization: being occurred more Connection situation, super-pixel are undersized, and single super-pixel is cut into multiple discontinuous super-pixel etc., these situations can pass through Enhance connectivity to solve.The connectivity of image is the prior art after enhancing segmentation, and main thought is: creating a label table, table Interior element is -1, according to " Z " type trend (from left to right, from top to bottom sequence) by discontinuous super-pixel, undersized super Pixel is reassigned to neighbouring super-pixel, and traversed pixel distributes to corresponding label, until all the points traversal finishes Until.

Claims (5)

1. a kind of adapting to image target area dividing method based on SLIC, which comprises the following steps:
Step 1: reading original image, down-sampling processing is carried out to the original image, down-sampling step-length is S pixel unit, i.e., All pixels point of the original image in the sliding window of S*S is merged into a super-pixel point, the institute in the super-pixel point Have in pixel, using the pixel that gray value is the average gray value of all pixels point in the super-pixel point as the super-pixel point Representative；
Step 2: the image seed point number after down-sampling is initialized, it, will according to the seed point number after initialization Seed point is evenly distributed in the various pieces of image；
Step 3: calculating the shade of gray of each pixel in the range of the n*n centered on seed point, gray scale is then found out The smallest pixel of gradient, using the position of the pixel as the final position of seed point；
Step 4: carrying out classification marker to the pixel around the seed point in step 4, specifically include:
Step 4 one scans for the pixel around seed point, and moving step length used in search process is Step= Sqrt (N/K)+q, wherein N is the quantity of pixel in original image, and K is the quantity of super-pixel point after segmentation, and q is positive integer；
Step 4 two, calculating each pixel in search range, at a distance from seed point, which includes Gray homogeneity d1, space Distance d2With induction Hash distance d3,
Wherein m, n are the pixel number in corresponding super-pixel block respectively, and k is the gray value of pixel；
Wherein x, y are the abscissa and ordinate of corresponding pixel points respectively；
Wherein, X indicates seed point and difference of the neighbouring super pixels point in gray scale on its up and down four direction；
Step 4 three, the total variances value D for calculating each pixel and seed point in search range:
D=μ1d1+μ2d2+μ3d3
Wherein μ1、μ2And μ3Respectively represent the weight coefficient of each distance, and μ1、μ2、μ3Meet following relationship:
μ1=μ2=5 μ3=log2K
Total variances value is compared step 4 four with according to preset threshold value, determines which cluster picture pixel belongs to this Vegetarian refreshments center；
Step 5: step 3 and step 4 is repeated several times, until cluster result does not change, image point is so far completed It cuts.
2. the method according to claim 1, wherein further include:
Step 6: enhancing the connectivity of image after segmentation.
3. the method according to claim 1, wherein the S=2 in step 1.
4. the method according to claim 1, wherein the q=2 in step 4 one.
5. the method according to claim 1, wherein the number of repetition in step 5 is 10.
CN201811258689.2A 2018-10-26 2018-10-26 A kind of adapting to image target area dividing method based on SLIC Pending CN109598726A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811258689.2A CN109598726A (en) 2018-10-26 2018-10-26 A kind of adapting to image target area dividing method based on SLIC

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811258689.2A CN109598726A (en) 2018-10-26 2018-10-26 A kind of adapting to image target area dividing method based on SLIC

Publications (1)

Publication Number Publication Date
CN109598726A true CN109598726A (en) 2019-04-09

Family

ID=65957391

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811258689.2A Pending CN109598726A (en) 2018-10-26 2018-10-26 A kind of adapting to image target area dividing method based on SLIC

Country Status (1)

Country Link
CN (1) CN109598726A (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110222644A (en) * 2019-06-10 2019-09-10 哈尔滨理工大学 Forest fire smoke detection method based on image segmentation
CN110443809A (en) * 2019-07-12 2019-11-12 太原科技大学 Structure sensitive property color images super-pixel method with boundary constraint
CN111833362A (en) * 2020-06-17 2020-10-27 北京科技大学 Unstructured road segmentation method and system based on superpixel and region growing
CN112669326A (en) * 2020-12-07 2021-04-16 山东省科学院海洋仪器仪表研究所 Rapid segmentation method for underwater image of seawater
CN113052859A (en) * 2021-04-20 2021-06-29 哈尔滨理工大学 Super-pixel segmentation method based on self-adaptive seed point density clustering
CN113344947A (en) * 2021-06-01 2021-09-03 电子科技大学 Super-pixel aggregation segmentation method
CN113838061A (en) * 2021-07-28 2021-12-24 中科云谷科技有限公司 Method and device for image annotation and storage medium
CN114140462A (en) * 2021-12-10 2022-03-04 江苏牛犇轴承有限公司 Bearing wear degree evaluation method based on image processing
CN115294159A (en) * 2022-09-30 2022-11-04 南通市博洋铸造有限公司 Method for dividing corroded area of metal fastener
CN116758059A (en) * 2023-08-10 2023-09-15 吉林交通职业技术学院 Visual nondestructive testing method for roadbed and pavement
CN117474897A (en) * 2023-11-21 2024-01-30 东莞市江霖五金实业有限公司 Visual detection method for surface defects of charger
CN117876646A (en) * 2024-03-11 2024-04-12 陕西仙喜辣木茯茶有限公司 Fuzhuan tea flowering image acquisition method
CN118097310A (en) * 2024-04-24 2024-05-28 陕西丰京建材有限公司 Method for digitally detecting concrete surface defects

Citations (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102129549A (en) * 2011-01-29 2011-07-20 广西师范大学 Image Hash method based on thumbnail and singular value decomposition
US20120250984A1 (en) * 2010-12-01 2012-10-04 The Trustees Of The University Of Pennsylvania Image segmentation for distributed target tracking and scene analysis
US20130188869A1 (en) * 2012-01-20 2013-07-25 Korea Advanced Institute Of Science And Technology Image segmentation method using higher-order clustering, system for processing the same and recording medium for storing the same
CN103353987A (en) * 2013-06-14 2013-10-16 山东大学 Superpixel segmentation method based on fuzzy theory
CN103761736A (en) * 2014-01-14 2014-04-30 宁波大学 Image segmentation method based on Bayes harmonious degree
CN104318259A (en) * 2014-10-20 2015-01-28 北京齐尔布莱特科技有限公司 Target picture identifying device and method for and computing device
US20150234863A1 (en) * 2014-02-18 2015-08-20 Environmental Systems Research Institute (ESRI) Automated feature extraction from imagery
US20150327766A1 (en) * 2014-05-19 2015-11-19 Ricoh Co., Ltd. Superpixel-based image segmentation using shading and albedo decomposition
CN105118049A (en) * 2015-07-22 2015-12-02 东南大学 Image segmentation method based on super pixel clustering
JP2016001404A (en) * 2014-06-11 2016-01-07 キヤノン株式会社 Image processing method, image processor and computer program
CN105354599A (en) * 2015-11-25 2016-02-24 安徽创世科技有限公司 Color recognition method based on improved SLIC super-pixel segmentation algorithm
KR101611466B1 (en) * 2015-01-05 2016-04-11 한국기술교육대학교 산학협력단 Subsampling-based Clustering Method for fast superpixel segmentation
JP2016129005A (en) * 2015-01-05 2016-07-14 富士フイルム株式会社 Hash value generation method and device, and program
AU2014280957A1 (en) * 2014-12-24 2016-07-14 Canon Kabushiki Kaisha Video segmentation method
CN105787965A (en) * 2016-01-26 2016-07-20 安徽创世科技股份有限公司 Image searching method based on color features
CN106056159A (en) * 2016-06-03 2016-10-26 西安电子科技大学 Image fine classification method based on Fisher Vector
CN106250895A (en) * 2016-08-15 2016-12-21 北京理工大学 A kind of remote sensing image region of interest area detecting method
US20170053182A1 (en) * 2014-04-29 2017-02-23 Institute of Automation Chinese Academy of Science Image matching method based on cascaded binary encoding
CN106815842A (en) * 2017-01-23 2017-06-09 河海大学 A kind of improved image significance detection method based on super-pixel
CN106952271A (en) * 2017-03-08 2017-07-14 西北工业大学 A kind of image partition method handled based on super-pixel segmentation and EM/MPM
CN107093205A (en) * 2017-03-15 2017-08-25 北京航空航天大学 A kind of three dimensions building window detection method for reconstructing based on unmanned plane image
CN107123130A (en) * 2017-03-06 2017-09-01 华南理工大学 Kernel correlation filtering target tracking method based on superpixel and hybrid hash
CN107230206A (en) * 2017-06-02 2017-10-03 太原理工大学 A kind of 3D Lung neoplasm dividing methods of the super voxel sequence lung images based on multi-modal data
CN108090894A (en) * 2017-11-24 2018-05-29 浙江理工大学 Based on the fabric defect detection method apart from adaptation function and perception hash algorithm
US20180278957A1 (en) * 2015-09-25 2018-09-27 Sisvel Technology S.R.L. Methods and Apparatuses for Encoding and Decoding Digital Images Through Superpixels

Patent Citations (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120250984A1 (en) * 2010-12-01 2012-10-04 The Trustees Of The University Of Pennsylvania Image segmentation for distributed target tracking and scene analysis
CN102129549A (en) * 2011-01-29 2011-07-20 广西师范大学 Image Hash method based on thumbnail and singular value decomposition
US20130188869A1 (en) * 2012-01-20 2013-07-25 Korea Advanced Institute Of Science And Technology Image segmentation method using higher-order clustering, system for processing the same and recording medium for storing the same
CN103353987A (en) * 2013-06-14 2013-10-16 山东大学 Superpixel segmentation method based on fuzzy theory
CN103761736A (en) * 2014-01-14 2014-04-30 宁波大学 Image segmentation method based on Bayes harmonious degree
US20150234863A1 (en) * 2014-02-18 2015-08-20 Environmental Systems Research Institute (ESRI) Automated feature extraction from imagery
US20170053182A1 (en) * 2014-04-29 2017-02-23 Institute of Automation Chinese Academy of Science Image matching method based on cascaded binary encoding
US20150327766A1 (en) * 2014-05-19 2015-11-19 Ricoh Co., Ltd. Superpixel-based image segmentation using shading and albedo decomposition
JP2016001404A (en) * 2014-06-11 2016-01-07 キヤノン株式会社 Image processing method, image processor and computer program
CN104318259A (en) * 2014-10-20 2015-01-28 北京齐尔布莱特科技有限公司 Target picture identifying device and method for and computing device
AU2014280957A1 (en) * 2014-12-24 2016-07-14 Canon Kabushiki Kaisha Video segmentation method
JP2016129005A (en) * 2015-01-05 2016-07-14 富士フイルム株式会社 Hash value generation method and device, and program
KR101611466B1 (en) * 2015-01-05 2016-04-11 한국기술교육대학교 산학협력단 Subsampling-based Clustering Method for fast superpixel segmentation
CN105118049A (en) * 2015-07-22 2015-12-02 东南大学 Image segmentation method based on super pixel clustering
US20180278957A1 (en) * 2015-09-25 2018-09-27 Sisvel Technology S.R.L. Methods and Apparatuses for Encoding and Decoding Digital Images Through Superpixels
CN105354599A (en) * 2015-11-25 2016-02-24 安徽创世科技有限公司 Color recognition method based on improved SLIC super-pixel segmentation algorithm
CN105787965A (en) * 2016-01-26 2016-07-20 安徽创世科技股份有限公司 Image searching method based on color features
CN106056159A (en) * 2016-06-03 2016-10-26 西安电子科技大学 Image fine classification method based on Fisher Vector
CN106250895A (en) * 2016-08-15 2016-12-21 北京理工大学 A kind of remote sensing image region of interest area detecting method
CN106815842A (en) * 2017-01-23 2017-06-09 河海大学 A kind of improved image significance detection method based on super-pixel
CN107123130A (en) * 2017-03-06 2017-09-01 华南理工大学 Kernel correlation filtering target tracking method based on superpixel and hybrid hash
CN106952271A (en) * 2017-03-08 2017-07-14 西北工业大学 A kind of image partition method handled based on super-pixel segmentation and EM/MPM
CN107093205A (en) * 2017-03-15 2017-08-25 北京航空航天大学 A kind of three dimensions building window detection method for reconstructing based on unmanned plane image
CN107230206A (en) * 2017-06-02 2017-10-03 太原理工大学 A kind of 3D Lung neoplasm dividing methods of the super voxel sequence lung images based on multi-modal data
CN108090894A (en) * 2017-11-24 2018-05-29 浙江理工大学 Based on the fabric defect detection method apart from adaptation function and perception hash algorithm

Non-Patent Citations (10)

* Cited by examiner, † Cited by third party
Title
LONG MA ET AL.: "Region-of-Interest Detection via Superpixel-to-Pixel Saliency Analysis for Remote Sensing Image", 《IEEE GEOSCIENCE AND REMOTE SENSING LETTERS》 *
XINGPING DONG ET AL.: "Hierarchical Superpixel-to-Pixel Dense Matching", 《IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY》 *
刘靖逸 等: "基于超像素分类的海天线检测方法", 《计量与测试技术》 *
吕晓佳: "基于多阈值的自适应SAR图像分割算法研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 *
周治平 等: "基于超像素的目标跟踪方法研究", 《光电工程》 *
姜金祎: "基于感知哈希编码的指纹与指静脉双模态识别方法研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 *
张小凤等: "基于图像超像素分析的图像分割方法", 《计算机技术与发展》 *
李啸宇 等: "融合局部纹理特征的颗粒图像SLIC超像素分割方法", 《食品与机械》 *
李晓媛: "基于区域合并的监督SAR图像分割算法研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 *
覃发超: "极化SAR影像超像素分割和面向对象分类方法研究", 《中国优秀博硕士学位论文全文数据库(博士)基础科学辑》 *

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110222644A (en) * 2019-06-10 2019-09-10 哈尔滨理工大学 Forest fire smoke detection method based on image segmentation
CN110443809A (en) * 2019-07-12 2019-11-12 太原科技大学 Structure sensitive property color images super-pixel method with boundary constraint
CN111833362A (en) * 2020-06-17 2020-10-27 北京科技大学 Unstructured road segmentation method and system based on superpixel and region growing
CN112669326A (en) * 2020-12-07 2021-04-16 山东省科学院海洋仪器仪表研究所 Rapid segmentation method for underwater image of seawater
CN113052859A (en) * 2021-04-20 2021-06-29 哈尔滨理工大学 Super-pixel segmentation method based on self-adaptive seed point density clustering
CN113344947A (en) * 2021-06-01 2021-09-03 电子科技大学 Super-pixel aggregation segmentation method
CN113838061A (en) * 2021-07-28 2021-12-24 中科云谷科技有限公司 Method and device for image annotation and storage medium
CN114140462B (en) * 2021-12-10 2023-09-08 江苏牛犇轴承有限公司 Bearing wear degree assessment method based on image processing
CN114140462A (en) * 2021-12-10 2022-03-04 江苏牛犇轴承有限公司 Bearing wear degree evaluation method based on image processing
CN115294159A (en) * 2022-09-30 2022-11-04 南通市博洋铸造有限公司 Method for dividing corroded area of metal fastener
CN116758059A (en) * 2023-08-10 2023-09-15 吉林交通职业技术学院 Visual nondestructive testing method for roadbed and pavement
CN116758059B (en) * 2023-08-10 2023-10-20 吉林交通职业技术学院 Visual nondestructive testing method for roadbed and pavement
CN117474897A (en) * 2023-11-21 2024-01-30 东莞市江霖五金实业有限公司 Visual detection method for surface defects of charger
CN117474897B (en) * 2023-11-21 2024-05-28 东莞市江霖五金实业有限公司 Visual detection method for surface defects of charger
CN117876646A (en) * 2024-03-11 2024-04-12 陕西仙喜辣木茯茶有限公司 Fuzhuan tea flowering image acquisition method
CN117876646B (en) * 2024-03-11 2024-05-28 陕西仙喜辣木茯茶有限公司 Fuzhuan tea flowering image acquisition method
CN118097310A (en) * 2024-04-24 2024-05-28 陕西丰京建材有限公司 Method for digitally detecting concrete surface defects
CN118097310B (en) * 2024-04-24 2024-07-09 陕西丰京建材有限公司 Method for digitally detecting concrete surface defects

Similar Documents

Publication Publication Date Title
CN109598726A (en) A kind of adapting to image target area dividing method based on SLIC
CN110909800B (en) Vehicle detection method based on Faster R-CNN improved algorithm
CN103810744B (en) It is backfilled a little in cloud
WO2017113232A1 (en) Product classification method and apparatus based on deep learning
JP5526955B2 (en) Face clustering device, face clustering method, and program
CN108470354A (en) Video target tracking method, device and realization device
CN105956560A (en) Vehicle model identification method based on pooling multi-scale depth convolution characteristics
Xu et al. Stereo computation using radial adaptive windows
JP2005284487A (en) Information processor and method, recording medium, and program
CN108154158B (en) Building image segmentation method for augmented reality application
JP6897749B2 (en) Learning methods, learning systems, and learning programs
CN103679187B (en) Image-recognizing method and system
CN110188763B (en) Image significance detection method based on improved graph model
CN103996052A (en) Three-dimensional face gender classification device and method based on three-dimensional point cloud
CN110751027B (en) Pedestrian re-identification method based on deep multi-instance learning
CN112949338A (en) Two-dimensional bar code accurate positioning method combining deep learning and Hough transformation
CN109492645A (en) A kind of registration number character dividing method and device
JPWO2015136709A1 (en) Image processing apparatus, image sensor, and image processing method
CN106600610A (en) FCM image segmentation method and apparatus
CN116079749A (en) Robot vision obstacle avoidance method based on cluster separation conditional random field and robot
CN108549915A (en) Image hash code training pattern algorithm based on two-value weight and classification learning method
CN113496260A (en) Grain depot worker non-standard operation detection method based on improved YOLOv3 algorithm
CN111612099B (en) Texture image classification method and system based on local sorting difference refinement mode
CN113327269A (en) Unmarked cervical vertebra movement detection method
CN111368674B (en) Image recognition method and device

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20190409

RJ01 Rejection of invention patent application after publication