Image Segmentation Based on Model Selection
Haojun Sun · 2006
Clustering-based image segment approach is popular in image processing. It consists in separating pixel features into clusters representing homogeneous regions. In the kind of methods, determining the number of clusters is an open problem. In this paper, we propose an efficient model selection algorithm for automatically determining the number of clusters. The algorithm roots the try-and-error approach. Due to the previous results to be used for initializing the next trail procedure, the FCM converge fast. So, the proposed algorithm is more efficient in term of computational time. Experimental results confirm efficiency of the proposed algorithm