Incorporating Feature Based Priors into the Geodesic Active Contour Model and its Application in Biomedical Imagery

Abstract

This paper presents improvements to the geodesic active contour (GAC) model obtained by incorporating user defined prior information into the model itself. Specifically, the stopping function in the GAC model is revised by designing an indicator function derived from a-priori information. The numerical implementation is based on the level set technique. Experimental results illustrate that our approach is efficient and feasible for both artificial and real images. In particular, the proposed method performs well in situations where existing methods are known to fail.