A novel SOM-based approach for active contour modeling

Abstract

We integrate the advantages of SOM- and snake-based ACMs in order to extract the desired contour from images. We employ: (i) the feature points to guide the contour, as in the case of SOM-based ACMs; (ii) the gradient and intensity variations in a local region to control the contour movement. However, in contrast with the snake-based ACMs, we do not use an explicit energy functional based on gradient or intensity. The algorithm is tested on synthetic binary and gray-level images, and the results show the superiority of the proposed algorithm over other conventional SOM- and snake-based ACM algorithms.