New local region-based model for inhomogeneous images with the simplex constraints of landmarks

Abstract

Image segmentation is widely used in different fields, like the medical field, security, and defense, etc. There are different techniques used for image segmentation some of these are the Mumford-Shah model, Chan-Vese (CV) model and the Local Gaussian Distribution Fitting (LGDF) model. But in the case of the images having weak edges and severe noise segmentation is still a difficult task. In this study, we will work on the segmentation of images having severe noise and weak boundaries. This is done by introducing landmark constraints in local Gaussian distribution fitting energy. Landmarks highlight the prominent features of an image. Finally, comparison is given in both qualitative and quantitative ways.