Object contour detection using an improved snake model

Abstract

This paper proposes an improved snake algorithm associated with new edge proximity energy for locating object contours robustly. The proposed algorithm represents the distance between the current snake point and its nearby edge as a new energy term and adds the term to the total energy function so that object contours can be extracted more accurately. The suggested snake model can not only detect object boundaries which are convex and concave but also detect the boundaries of complex-shape objects without heuristic adjustment of weights given to energy terms. Experimental results show that our snake algorithm can extract object boundaries more successfully than other existing methods. ©2013 International Information Institute.