Abstract: Interactive image segmentation methods usually require users to provide much more initial seeds or more than once interactive when they are used for medical image segmentation with fuzzy boundaries．This paper presented an optimized interactive image segmentation algorithm with multi-thresholds technology．The proposed algorithm is based on GorwCut algorithm and improved in two aspects:one is automatically generating initial seeds templates by image gray histogram with multi-thresholds,and the other is raising iterative efficiency by improved cellular automaton iterative algorithm．Compared with GrowCut algorithm,the proposed algorithm simplifies the user interactive operations and improves the segmentation accuracy．Experimental results on clinical plaque and liver image segmentations demostrate the sound performances of the proposed algorithm.

Key words: Interactive,Multi-thresholds,Histogram,Cellular automata,Medical image segmentation