Edge-Based Segmentation Algorithm for SAR Satellite Images Using Optimized Fuzzy Cellular Automata

Recently, great advances have been made to improve resolution of synthetic aperture radar (SAR) images. Numerous studies have been conducted on segmentation of SAR images for developing and updating geographical information. The present paper proposes a fuzzy cellular automata algorithm for segmenting SAR images. This theory is a combination of cellular automata and fuzzy rules to develop a model called fuzzy cellular automata. In this paper, 8-tuple Moore neighborhood has been used. Three variables including primary cells, immune status of neighboring cells and gray-level of neighboring cells were investigated in eight different directions Moore and fuzzy rules have been imposed on them. Then, the results have been tested on two sets of images. The first is a set of optical and simulated SAR images. The images were first segmented by FCA algorithm. Then, the results were compared with the results of GA and AFS algorithms. The error rates of previous FCA and the proposed FCA have been calculated. The second set includes real SAR images. After segmenting by FCA algorithm, the results were compared with the results of GA, AFS, ABC, histogram-based thresholding algorithm and previous FCA algorithms. The results show that the proposed method is more appropriate for segmenting SAR images because of higher accuracy and less error rate than other methods.