Automatic image segmentation incorporating shape priors via graph cuts

Abstract

In recent years, graph cut has been regarded as an effective discrete optimization method and received increasing attentions in vision community. However, many existing graph cut segmentation algorithms require interactive operations, which are not appropriate for automatic applications. In this paper, we propose an automatic segmentation algorithm via graph cut. Firstly, the data term in traditional graph cut energy is redefined to counteract illumination change. Secondly, shape priors are introduced into segmentation process, which help to obtain more robust results. Finally, an automatic segmentation strategy is presented. Experiments demonstrate that our segmentation algorithm can provide promising results, even when object suffering pixel intensity variation and continuously shape deformation.