Fast image segmentation by convex minimisation and split Bregman method

Abstract

A convex minimisation model for image segmentation is proposed. The basic idea of this model is that objects will be detected automatically if background is removed. The local information of every pixel is used to make the model applicable to images with intensity inhomogeneity. Also, by using a convex approximation of the Heaviside function, the convex energy function of the proposed model is obtained. Then it is minimised by applying the split Bregman method, which is a fast technique to obtain the global minimiser. The experimental results demonstrate that the proposed model is powerful in efficiency and accuracy.