Texture-aware fast global level set evolution

Abstract

Due to its intrinsic advantages such as the ability to automatically handle complex shapes and topological changes, the level set method has been widely used in image segmentation. Nevertheless, in addition to be computational expensive, it has the limitation to very often lead to a local minimum because of the energy functional to be minimized is non-convex. In this work, we use the geometric active contours and the image thresholding frameworks to design a novel method for global image segmentation. The local lattice Boltzmann method is used to solve the level set equation. The proposed algorithm is therefore effective and highly parallelizable. Experimental results on satellite, natural and medical images demonstrate the effectiveness and the efficiency of the proposed method when implemented using an NVIDIA graphics processing units. © 2013 Springer-Verlag Berlin Heidelberg.