Research Of Image Segmentation Based On Fuzzy Correlation And Wave Cluster

The image segmentation algorithm based on Wave Cluster has been widely concerned because of its advantages of fast running speed and effective extraction of target regions.However,the Wave Cluster algorithm has the disadvantage that it is difficult to determine the division threshold and it has low clustering accuracy.The image segmentation results through Wave Cluster have problems of partial volume effect and isolated noise.Therefore,this paper first makes improvements on Wave Cluster.In the improved Wave Cluster,when the clusters found in the wavelet domain are mapped back to the grid space,the grid cells at the edge of the clusters are merged into the clusters according to the similarity between the grids cells.Then,the maximum fuzzy correlation is introduced into Wave Cluster algorithm.The fuzzy correlation Wave Cluster algorithm is proposed to make up the lack of dependence on the division threshold in Wave Cluster.Finally,the fuzzy correlation algorithm and Wave Cluster algorithm are introduced into the field of color image segmentation,and an image segmentation algorithm based on fuzzy correlation and Wave Cluster(FCWC)is proposed.First,in order to achieve the purpose of eliminating the texture of the image and highlighting the edge of the image,the main structure is extracted.Then,in order to solve the problem of partial volume effect caused by Wave Cluster in the image segmentation process,the main structure image is divided into several superpixels.Subsequently,we simplified the color characteristics of the superpixels and calculated weights of superpixels.Finally,the fuzzy correlation Wave Cluster algorithm is used to obtain the cluster label.And assign labels to superpixels to get the final image segmentation result.The image segmentation algorithm proposed in this paper solves the problems of partial volume effect and isolated noise in the image segmentation results while ensuring segmentation efficiency.Experiments were conducted in the BSDS500 database.From the visual segmentation results,it can be seen that image segmentation algorithms based on fuzzy correlation and Wave Cluster can effectively segment color images.The segmentation results show that FCWC are superior to other comparison algorithms.Quantitative evaluation of the segmentation results of this algorithm and other comparison algorithms shows that the algorithm of this paper has the highest PRI,and the lowest Vo I,GCE,BDE,has the highest segmentation quality.There are 13 figure,4 tables and 65 reference in this paper.