Optimization of Level Set Segmentation with Nonlinear Coherent Diffusion

Abstract

Level set method is convenient in image segmentation for the stabilization and veracity. Gaussian filter is usually taken as a preprocess to reduce the influence of weak edges due to noises, but the disadvantage is obvious: blur fine structures specially the important boundaries and lead to inaccurate segmentation result. This paper introduces a robust method which filters the images with a Nonlinear Coherent Diffusion (NCD) to accelerate the evolution of level set in a spatially varying manner. Experimental results show the performance of the proposed method in improving precision of segmentation.