Dynamic Fuzzy Local Information C-Mean Clustering

Fuzzy local information c-mean (FLICM) algorithm is an image clustering that not only considers the intensity of gray level of pixels but also consider the spatial local information in the clastering process. This paper presents a new FLICM that in addition to all advantages of usual FLICM, it has a higher convergence speed. This is due to a dynamic operation of the algorithm. Some of novelties proposed in the algorithm are as follows: At the first, the algorithm uses a low resolution image which is obtained from the initial image and gradually increases the resolution in few stages. The convergence termination criterion for each resolution is different. The higher resolution is used, the lower error for convergence is considered. The window used for incorporating spatial local information has not a fixed size. In the beginning, the window has small size and gradually it becomes larger. The results of experiments on some images show that without loss of accuracy, the speed of algorithm has been increased

Keywords:

image clustering , image segmentation , dynamic clustering , fuzzy c-mean , local information