CN108765429A - A kind of image segmentation system based on clustering - Google Patents

A kind of image segmentation system based on clustering Download PDF

Info

Publication number
CN108765429A
CN108765429A CN201810482000.8A CN201810482000A CN108765429A CN 108765429 A CN108765429 A CN 108765429A CN 201810482000 A CN201810482000 A CN 201810482000A CN 108765429 A CN108765429 A CN 108765429A
Authority
CN
China
Prior art keywords
cluster
image
pixel
classification
module
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.)
Withdrawn
Application number
CN201810482000.8A
Other languages
Chinese (zh)
Inventor
杨金源
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.)
Shenzhen Zhida Machinery Technology Co Ltd
Original Assignee
Shenzhen Zhida Machinery Technology Co Ltd
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 Shenzhen Zhida Machinery Technology Co Ltd filed Critical Shenzhen Zhida Machinery Technology Co Ltd
Priority to CN201810482000.8A priority Critical patent/CN108765429A/en
Publication of CN108765429A publication Critical patent/CN108765429A/en
Withdrawn legal-status Critical Current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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Image Analysis (AREA)

Abstract

A kind of image segmentation system based on clustering, including image input module, image characteristics extraction module, Cluster Analysis module, image segmentation module and information display module, described image input module is for inputting original image to be split, the characteristic extracting module is used to extract the geometric properties and color characteristic of pixel in the original image, composition characteristic Vector Groups, the Cluster Analysis module is used to carry out cluster operation to each pixel in image according to described eigenvector group and evaluate cluster result, described image segmentation module is for being split described image according to cluster result, described information display module is used to show the image after cluster result and segmentation.Beneficial effects of the present invention are：Cluster operation is carried out to image to be split using improved PCM algorithms, the defect that PCM algorithms are easy to cause cluster result consistency is effectively overcome, obtains preferable cluster result so that the segmentation result of image more corresponds to actual needs.

Description

A kind of image segmentation system based on clustering
Technical field
The invention is related to image processing field, and in particular to a kind of image segmentation system based on clustering.
Background technology
In image processing field, how to image carry out effectively, accurate segmentation be puzzlement people always problem it One, in recent years, fuzzy theory, which is applied to image segmentation field, has become hot spot, because image is true with ambiguity and not It is qualitative, and fuzzy theory just has this ambiguity of processing and probabilistic inherent advantage.Image based on fuzzy theory Partitioning algorithm has very much, wherein foremost algorithm will belong to Fuzzy C-Means Cluster Algorithm (FCM algorithms), but FCM algorithms are to making an uproar The interference of sound is very sensitive and has ignored influence of the noise spot to cluster centre, in order to overcome these deficiencies, Krishnapuram et al. proposes a kind of PCM clustering algorithms, which can effectively cluster comprising noise or outlier Data, but PCM algorithms are very sensitive to initial cluster center, are easily trapped into local optimum, and cluster result consistency etc. is caused to be asked Topic.
The present invention provides a kind of image segmentation algorithm based on clustering, by the object function in PCM algorithms, person in servitude The problems such as category degree more new formula and Cluster Validity are evaluated is improved, and effectively overcomes PCM algorithms and is easy to cause cluster knot The defect of fruit consistency can obtain preferable cluster result, can not only effectively divide image, and make point of image Result is cut more to correspond to actual needs.
Invention content
In view of the above-mentioned problems, the present invention is intended to provide a kind of image segmentation system based on clustering.
The purpose of the invention is achieved through the following technical solutions：
A kind of image segmentation system based on clustering, including image input module, image characteristics extraction module, cluster Analysis module, image segmentation module and information display module, described image input module are used to input original image to be split, The characteristic extracting module is used to extract the geometric properties and color characteristic of pixel in the original image, to composition characteristic Vector Groups, the Cluster Analysis module include cluster arithmetic element and Cluster Assessment unit, and the cluster arithmetic element is used for root Each pixel that the feature vector group obtained according to extraction is treated in segmentation image carries out cluster operation, obtains cluster result, institute It states Cluster Assessment unit to evaluate obtained cluster result according to Cluster Validity Index, to judge to cluster operation result Quality, described image segmentation module is used to be split the image to be split according to the obtained result of cluster, the letter Cease evaluation result of the display module for real-time display cluster operation and the image after segmentation.
The advantageous effect of the invention：The present invention provides a kind of image segmentation algorithm based on clustering, by right The problems such as object function, degree of membership more new formula and Cluster Validity Index in PCM algorithms, is improved, and effectively overcomes PCM algorithms are easy to cause the defect of cluster result consistency, obtain preferable cluster result, can not only effective segmentation figure Picture, and the segmentation result of image is more corresponded to actual needs.
Description of the drawings
Innovation and creation are described further using attached drawing, but the embodiment in attached drawing does not constitute and appoints to the invention What is limited, for those of ordinary skill in the art, without creative efforts, can also be according to the following drawings Obtain other attached drawings.
Fig. 1 is schematic structural view of the invention；
Reference numeral：
Image input module 1；Image characteristics extraction module 2；Cluster Analysis module 3；Image segmentation module 4；Presentation of information Module 5；Cluster arithmetic element 31；Cluster Assessment unit 32.
Specific implementation mode
The invention will be further described with the following Examples.
Referring to Fig. 1, a kind of image segmentation system based on clustering of the present embodiment, including image input module 1, figure As characteristic extracting module 2, Cluster Analysis module 3 and image segmentation module 4 and information display module 5, described image input module 1 For inputting original image to be split, the characteristic extracting module 2 is used to extract the geometry of pixel in the original image Feature and color characteristic, to composition characteristic Vector Groups, the Cluster Analysis module 3 includes that cluster arithmetic element 31 and cluster are commented Valence unit 32, the cluster arithmetic element 31 are used to treat each picture in segmentation image according to the feature vector group that extraction obtains Vegetarian refreshments carries out cluster operation, obtains cluster result, and the Cluster Assessment unit 32 is poly- to what is obtained according to Cluster Validity Index Class result is evaluated, to judge to cluster the quality of operation, the result that described image segmentation module 4 is used to obtain according to cluster The image to be split is split, described information display module 5 clusters the evaluation result of operation for real-time display and divides Image after cutting.
Preferably, the geometric properties include the profile, texture and continuity of pixel, and the color characteristic includes pixel The color of point and brightness.
This preferred embodiment provides a kind of image segmentation algorithm based on clustering, is treated using improved PCM algorithms The image of segmentation carries out cluster operation, effectively overcomes the defect that PCM algorithms are easy to cause cluster result consistency, obtains Preferable cluster result can not only effectively divide image, and the segmentation result of image is made more to meet practical need It wants.
Preferably, cluster arithmetic element 31 carries out cluster operation using PCM algorithms to feature vector group, if set of pixels is X ={ x1,x2,…,xn, n is the pixel sum of image to be split, and c is the classification number of cluster, vi(i=1,2 ..., c) it is class The cluster centre of other i, m and uijIt is the Weighted Index sum number pixel x defined according to FCM algorithms respectivelyiClassification j is subordinate to Degree, k and wijIt is the Weighted Index defined according to PCM algorithms and pixel x respectivelyiTo the probability of classification j, then the PCM mesh that uses Scalar functions JPCM' be：
In formula, d (xj,vi) it is pixel xjTo cluster centre viEuclidean distance,It is the cluster of entire set of pixels Center,It is cluster centre viTo cluster centreEuclidean distance, vpAnd vqIt is the poly- of classification p and classification q respectively Class center, d (vp,vq) it is cluster centre vpTo cluster centre vqEuclidean distance.
Preferably, cluster arithmetic element 31 carries out the feature vector group in original image to be split using PCM algorithms It clusters, then the object function J of clustering algorithmPCM' in wijAnd uijCalculation formula be：
In formula, m is the Weighted Index defined according to FCM algorithms, and k is the Weighted Index defined according to PCM algorithms, d (xj, vi) it is pixel xjTo the cluster centre v of classification iiEuclidean distance,It is cluster centre viTo the poly- of set of pixels Class centerEuclidean distance, d (vp,vq) be classification p cluster centre vpTo the cluster centre v of classification qqEuclid Distance, d (xj,vl) it is pixel xjTo the cluster centre v of classification llEuclidean distance, c be cluster classification number, n is to wait for Divide the pixel sum of image.
This preferred embodiment clusters the image feature vector of acquirement using improved PCM algorithms, overcomes tradition PCM algorithms cluster consistency the problem of, in addition, making in practical applications, to greatly reduce this to the improvement of object function The complexity of algorithm, the preferable classifying quality not only gone, but also improve the efficiency of image segmentation system.
Preferably, the Cluster Assessment unit 32 is for evaluating the cluster result of gained, proposes a kind of new Cluster Validity Index P ' evaluates the Clustering Effect of the clustering algorithm, if n indicates pixel in image to be split sum, c Indicate that the classification number of cluster, k indicate Weighted Index, and k>1, viIndicate the cluster centre of classification i, vsIn the cluster for indicating classification s The heart, then the calculation formula of P ' be：
In formula, A and B are respectively coefficient, and A+B=1.
The value of Validity Index P ' is smaller, shows that the result of cluster is better, when P ' acquirement minimum values, i.e., corresponding best poly- Class result.
This preferred embodiment proposes a kind of new Cluster Validity Index to evaluate cluster result, passes through all clusters Class between the degree of coupling and class the sum of separating degree come define cluster result division quality so that new Cluster Validity Index is got over It is small to represent that Clustering Effect is better, when new Cluster Validity Index minimum, that is, best Clustering Effect is represented, can treated point The cluster result for cutting the pixel of image carries out effectively evaluating.
Finally it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than the present invention is protected The limitation of range is protected, although being explained in detail to the present invention with reference to preferred embodiment, those skilled in the art answer Work as understanding, technical scheme of the present invention can be modified or replaced equivalently, without departing from the reality of technical solution of the present invention Matter and range.

Claims (5)

1. a kind of image segmentation system based on clustering, characterized in that including image input module, image characteristics extraction mould Block, Cluster Analysis module, image segmentation module and information display module, described image input module is for inputting original to be split Beginning image, the characteristic extracting module are used to extract the geometric properties and color characteristic of pixel in the original image, to Composition characteristic Vector Groups, the Cluster Analysis module include cluster arithmetic element and Cluster Assessment unit, the cluster operation list Each pixel that feature vector group of the member for being obtained according to extraction is treated in segmentation image carries out cluster operation, is clustered As a result, the Cluster Assessment unit evaluates obtained cluster result according to Cluster Validity Index, to judge to cluster The quality of operation result, described image segmentation module are used to divide the image to be split according to the result that cluster obtains It cuts, evaluation result of the described information display module for real-time display cluster operation and the image after segmentation.
2. a kind of image segmentation system based on clustering according to claim 1, characterized in that the geometric properties Profile, texture including pixel and continuity, the color characteristic include color and the brightness of pixel.
3. a kind of image segmentation system based on clustering according to claim 2, characterized in that cluster arithmetic element Cluster operation is carried out to feature vector group using PCM algorithms, if set of pixels is X={ x1,x2,…,xn, n is image to be split Pixel sum, c are the classification number of cluster, vi(i=1,2 ..., c) is the cluster centre of classification i, m and uijIt is basis respectively The Weighted Index and pixel x that FCM algorithms defineiTo the degree of membership of classification j, k and wijIt is to be added according to what PCM algorithms defined respectively Weigh index and pixel xiTo the probability of classification j, then the object function J of the PCM algorithms usedPCM' be：
In formula, d (xj,vi) it is pixel xjTo cluster centre viEuclidean distance,It is the cluster centre of entire set of pixels,It is cluster centre viTo cluster centreEuclidean distance, vpAnd vqBe respectively classification p and classification q cluster in The heart, d (vp,vq) it is cluster centre vpTo cluster centre vqEuclidean distance.
4. a kind of image segmentation system based on clustering according to claim 3, characterized in that cluster arithmetic element The feature vector group in original image to be split is clustered using PCM algorithms, then the object function J of clustering algorithmPCM′ In wijAnd uijCalculation formula be：
In formula, m is the Weighted Index defined according to FCM algorithms, and k is the Weighted Index defined according to PCM algorithms, d (xj,vi) be Pixel xjTo the cluster centre v of classification iiEuclidean distance,It is cluster centre viTo the cluster centre of set of pixelsEuclidean distance, d (vp,vq) be classification p cluster centre vpTo the cluster centre v of classification qqEuclidean distance, d (xj,vl) it is pixel xjTo the cluster centre v of classification llEuclidean distance, c be cluster classification number, n is figure to be split The pixel sum of picture.
5. a kind of image segmentation system based on clustering according to claim 4, characterized in that Cluster Assessment unit It is evaluated for the cluster result to gained, proposes a kind of new Cluster Validity Index P ' to evaluate the clustering algorithm Clustering Effect, if n indicates that the sum of the pixel in image to be split, c indicate that the classification number of cluster, k indicate Weighted Index, and k> 1, viIndicate the cluster centre of classification i, vsIndicate the cluster centre of classification s, then the calculation formula of P ' is：
In formula, A and B are respectively coefficient, and A+B=1.
The value of Validity Index P ' is smaller, shows that the result of cluster is better, when P ' acquirement minimum values, i.e., corresponding best cluster knot Fruit.
CN201810482000.8A 2018-05-18 2018-05-18 A kind of image segmentation system based on clustering Withdrawn CN108765429A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810482000.8A CN108765429A (en) 2018-05-18 2018-05-18 A kind of image segmentation system based on clustering

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810482000.8A CN108765429A (en) 2018-05-18 2018-05-18 A kind of image segmentation system based on clustering

Publications (1)

Publication Number Publication Date
CN108765429A true CN108765429A (en) 2018-11-06

Family

ID=64008389

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810482000.8A Withdrawn CN108765429A (en) 2018-05-18 2018-05-18 A kind of image segmentation system based on clustering

Country Status (1)

Country Link
CN (1) CN108765429A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112954261A (en) * 2021-03-18 2021-06-11 深圳奇实科技有限公司 Video conference network flow control method and system
CN113011221A (en) * 2019-12-19 2021-06-22 广州极飞科技股份有限公司 Crop distribution information acquisition method and device and measurement system

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113011221A (en) * 2019-12-19 2021-06-22 广州极飞科技股份有限公司 Crop distribution information acquisition method and device and measurement system
CN112954261A (en) * 2021-03-18 2021-06-11 深圳奇实科技有限公司 Video conference network flow control method and system
CN112954261B (en) * 2021-03-18 2021-09-10 深圳奇实科技有限公司 Video conference network flow control method and system

Similar Documents

Publication Publication Date Title
CN100361451C (en) System for detecting eroticism and unhealthy images on network based on content
CN107679183B (en) Training data acquisition method and device for classifier, server and storage medium
WO2018023734A1 (en) Significance testing method for 3d image
CN110428432A (en) The deep neural network algorithm of colon body of gland Image Automatic Segmentation
CN104504734B (en) A kind of color of image transmission method based on semanteme
US10986400B2 (en) Compact video representation for video event retrieval and recognition
CN108921820B (en) Saliency target detection method based on color features and clustering algorithm
CN107016409A (en) A kind of image classification method and system based on salient region of image
CN106296695A (en) Adaptive threshold natural target image based on significance segmentation extraction algorithm
CN105205804A (en) Caryon-cytolymph separation method and apparatus of white blood cells in blood cell image, and classification method and apparatus of white blood cells in blood cell image
CN110163239A (en) A kind of Weakly supervised image, semantic dividing method based on super-pixel and condition random field
CN101615245A (en) Expression recognition method based on AVR and enhancing LBP
CN112906550B (en) Static gesture recognition method based on watershed transformation
CN106960182A (en) A kind of pedestrian integrated based on multiple features recognition methods again
CN109360191B (en) Image significance detection method based on variational self-encoder
CN108615229A (en) Collision detection optimization method based on curvature points cluster and decision tree
CN108734200A (en) Human body target visible detection method and device based on BING features
CN112434647A (en) Human face living body detection method
CN108765429A (en) A kind of image segmentation system based on clustering
CN107169996A (en) Dynamic human face recognition methods in a kind of video
CN106203448A (en) A kind of scene classification method based on Nonlinear Scale Space Theory
CN115240119A (en) Pedestrian small target detection method in video monitoring based on deep learning
CN109934852B (en) Video description method based on object attribute relation graph
Kumar et al. A hybrid cluster technique for improving the efficiency of colour image segmentation
CN105550642B (en) Gender identification method and system based on multiple dimensioned linear Differential Characteristics low-rank representation

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
WW01 Invention patent application withdrawn after publication

Application publication date: 20181106

WW01 Invention patent application withdrawn after publication