Multiple seeded region growing algorithm for image segmentation using local extrema

In this paper, a multiple seeded region growing technique for image segmentation is presented. Conventional image segmentation techniques using region growing requires initial seeds selection, which increases computational cost and execution time. To overcome this problem, a seeded region growing technique for image segmentation is proposed, which starts from searching for local extrema of the image using morphology as the initial seeds, whose coordinates are saved in a pair of static FIFO queues, used for wave region growing. It grows regions according to the extreme values quasi-parallel. We use intensity based similarity index for the grow regions and adaptive threshold is used to calculate the criteria for the grow new waves. We apply the proposed algorithm to the Berkley segmentation dataset and discuss results based on F and SSIM evaluation functions that show efficient segmentation