Abstract: This paper proposes an automatic image segmentation algorithm which combines spatial clustering and edge gradient information. While evaluating the color and texture similarity of super pixel, a more accurate calculation method of segmented edge gradient is proposed. The geodesic distance is used to describe the similarity between super pixels, so that the segmentation results better integrate edge discontinuity and regional similarity in the image. A large number of experimental results of image segmentation show that this method can find the segmentation boundary more accurately and improve the accuracy of image segmentation.

Key words: image segmentation, gradient, super pixel, geodesic distance