A new image segmentation algorithm based on SRG and ACM

Bo Cai

Abstract

This paper proposes a new region-based active contour model for image segmentation, which is robust to background difference and intensity non-uniformity. The proposed model is different from other general regionbased models: Firstly, SRG model is added to the segmentation as a pre-segment process; Secondly, considering of intensity and gradient variation of the different input images and regions, we using the contour of pre-segmented region’s contour as the iteration source, it make the evolution be robust on condition of the noise and intensity nonuniformity; In addition, the gradient information and the pre-segmented edge information are combined to segment the region, this process enhances the ability of capturing the complex topological structures and making the edge more accurate and stable. Experiments on the natural images demonstrated the desired segmentation performance of our proposed model for the image with intensity non-uniformity.