We analyzed the principle of the traditional local binary fitting operation, Gaussian kernel function weighted summation (GKFWS), to develop a novel level set model in this paper. In this model, the traditional GKFWS operation is replaced with the median filter operation in the second procedure of local fitting of the energy domain. Furthermore, we incorporated the edge stopping function of GAC model into it to introduce the edge information for segmentation. Experiments on synthetic and real images demonstrate that this model has promising performance in terms of computational cost, robustness to noises and segmentation of images with intensity inhomogeneity.
Keywords: Active contour; edge information; image segmentation; level set; median filter.