FUZZY SELECTING LOCAL REGION LEVEL SET ALGORITHM

In this work, we introduced a novel localized region based level set model which is simultaneously effective for heterogeneous object or/and background and robust against noise. As such, we propose to minimize an energy functional based on a selective local average, i.e., when computing the local average, instead to use the intensity of all the pixels surrounding a given pixel, we first give a local Gaussian fuzzy membership to be a background or an object pixel to each of these surrounding pixels and then, we use the fuzzy weighted local average of these pixels to replace the traditional local average. With the graphics processing units' acceleration, the local lattice Boltzmann method is used to solve the proposed level set equation. The algorithm is effective in presence of intensity heterogeneity, robust against noise, fast and highly parallelizable. Experimental results demonstrate subjectively and objectively the performance of the proposed framework.