Boundary and region level set method based on local entropy for image segmentation

An improved level set method for image segmentation based on image local entropy information is proposed, which combines the edge-based model and the region-based model into a joint framework. Using image local entropy information, an adaptive weighting function is built firstly, which enables the evolving curve choose the evolution direction and move to the object boundary, adaptively. A novel edge indicator function is proposed based on the weighting function, which improves the ability of detecting weak boundary and accelerates the speed of contour evolution. Finally, the Chan-Vese (C-V) model is introduced into the joint framework as an external energy, which enhances the model dealing images with intensity inhomogeneity. In the experiments, the robustness of the method is evaluated to initial contours and noises, and the ability of segmenting images with intensity inhomogeneity. The results show that the proposed method can not only enhance the robustness to noises and improve the ability of segmenting images with weak boundary, but also achieve the satisfying results in segmenting images with intensity inhomogeneity, as compared with the other three methods using objective numerical indicators.