A novel selective segmentation model for images with intensity inhomogeneity

Abstract

Selective image segmentation is a very important and practical procedure in image processing. However, the existing selective segmentation model can not segment images with intensity inhomogeneity and fuzzy edge images accurately. In this paper a novel model which has better performance for images with intensity inhomogeneity and fuzzy edge images is proposed. The establishment of the model is based on the techniques of curve evolution, local statistic information and level set method. The stopping term in the new model with geometrical constraint is based on the LBF (Local Binary Fitting) active contour model. This paper applies the AOS (Additive Operator Splitting) method to speed up the segmentation. Experimental results on synthetic and real images clearly show that the novel model can segment the target more accurately, especially with respect to images with intensity inhomogeneity, weak edges and noise.