A Normalized Local Binary Fitting Model for Image Segmentation

Abstract

A normalized local binary fitting (NLBF) model is proposed for image segmentation in this paper. The proposed model can effectively and efficiently segment images with intensity in homogeneity because the image local characteristics are considered. At the same time, we use a Gaussian filtering process instead of the regularization to keep the level set function smooth in the evolution process. The strategy can reduce computational cost. Comparative experimental results on synthetic and real images demonstrate that the proposed model outperforms the well-known local binary fitting (LBF) model in computational efficiency and robustness to the initial contour.