A Novel Local Regional Model Based on Three-Layer Structure

Abstract

Translate Article

Abstract In this paper, considering the local variance of intensity inhomogeneity, we propose a novel local regional level set model based on a so-called Three-Layer structure to segment images with intensity inhomogeneity. The local region intensity mean idea is used to construct region descriptor. Especially, three descriptors separately based on ‘large’, ‘median’ and ‘small’ scales of local regions are utilized to derive the Three-Layer structure. Compared to the traditional methods based on fixed scale for all local regions, the Three-Layer structure is more reliable for capturing local intensity information. Then, the Three-Layer structure is incorporated into the level set energy functional construction. As a result, more effective local intensity information is incorporated into the level set evolution. Finally, the experimental results demonstrate that the proposed method yields results comparative to and even better than the existing popular models for segmenting images with intensity inhomogeneity.Keywordslocal informationlevel setintensity inhomogeneitythree-Layer structure