A nonlinear anisotropic diffusion model with non-standard growth for image segmentation

Abstract

The anisotropic diffusion equation with non-standard growth embodies the physical characteristics of “point-by-point anisotropy” as well as has important potential value in computer vision. In this paper, a general anisotropic diffusion framework of the level set function is proposed for image segmentation in scalar-value and vector-value images. Specifically, we develop a new regularization term that uses a diffusion coefficient with non-standard growth conditions and diffusion tensors. The existence and uniqueness of the model are obtained by the Galerkin method. We establish the numerical algorithms for obtaining the texture feature and evolving the level set function of images. Some numerical tests on medical and natural images confirm the accuracy of the proposed method and the improvement in segmenting small features.