Automatic local parameterization of the Chan Vese active contour model’s force coefficients using edge information

Abstract

Image segmentation is an important field of computer vision that includes algorithms that segment images into foreground and background. Active Contours (AC) is a family of segmentation techniques that has gained high popularity in recent years due to the advantages it provides, including contour deformation, detail preservation and multiple objects identification. One of the most popular AC algorithms is the Chan–Vese (CV) algorithm, which uses region information. This paper presents a novel algorithmic scheme named Automatic Local Ratio (ALR) that extends the original CV algorithm in order to automatically determine its force coefficient’s values, using edge information. Moreover, due to the use of edge information, this scheme provides an alternative to typical edge stopping functions. The proposed scheme is tested on real images, providing superior performance both in terms of qualitative and quantitative results compared to the CV using Gaussian regularization, the SBGFRLS and Fast Edge Integration algorithms.