Research Of Color Image Segmentation Algorithm Based On Watershed And Region Growing

Image segmentation is based on the consistency request of some feature to segment image into multiple areas.It is a key step in image analysis,but also a basis to understanding image better.Color image,because it can provide more information than gray level image,the segmentation of them has attracted more and more attentions.Currently,the available methods for color image segmentation are mostly expanded from gray-scale image segmentation.In this paper,the algorithm based on watershed and region growing is adopted;the small region removing algorithm is also used to removing the small regions after region growing.As follows:Firstly,watershed algorithm is used to forming closed segmentation regions.As its over-segmentation problem,the anisotropic P-M diffusion is adopted for smoothing preprocessing which can eliminate noise as well as maintain edge information,and so the problem of over-segmentation is effectively improved.Secondly,the automatic seed region selection is designed based on the initial segmentation regions,the seeded region is selected based on mean hue difference and the highest similarity is had among its adjacent regions.Thirdly,from these selected seed regions,the region growing algorithm is performed by inspecting the similarity of its adjacent regions.In order to reduce the number of small regions after region growing,the small region removing is used, and then the effective color image segmentation is achieved.The algorithm is simulated by MATLAB,also is compared with other color image segmentations,the experimental results show that the color image segmentation of this paper can get better segmentation results.