Segmentation of Colour Images by Modified Mountain Clustering

Defence Science Journal

M. Hanmandlu¹  Devendra Jha¹ 

[1]Scientific Analysis Group, Delhi

关键词: Fuzzy C-means clusters;    computer vision;    fuzzy Gaussian function;    fuzzy image p&;    cessing;    membership function;    validity criteria;    mountain clustering;   

DOI  :  

学科分类：社会科学、人文和艺术（综合）

来源: Defence Scientific Information & Documentation Centre

【 摘 要 】

Segmentation of colour images is animportant issue in various machine vision and image processingapplications. Though clustering techniques have been in vogue for manyyears, these have not been very effective because of problems likeselection of the number of clusters. This problem has been tackled byhaving a validity measure coupled with the new clustering technique.This method treats each point in the dataset, which is the map of allpossible colour combinations in the given image, as a potential clustercentre and estimates its potential wrt other data elements. First, thepoint with the maximum value of potential is considered to be a clustercentre and then its effect is removed from other points of the dataset.This procedure is repeated to determine different cluster centres. Atthe same time, the compactness and the minimum separation is computedamongst all the cluster centres, and also the validity function as theratio of these quantities. The validity function can be used to choosethe number of clusters. This technique has been compared to the fuzzyC-means technique and the results have been shown for a sample colourimage.

【 授权许可】

Unknown