Contour segmentation using an improved GAC model

An improved geodesic active contour model for the detection of object boundaries is presented. A geodesic active contour model is based on an active contour evolving in time. The geodesic active contour model has a serious weakness, that is, it stops at a local minimum where there is a deep concave boundary in an image. The generalized geodesic active contour model makes an active contour convergent to long, thin boundary indentations by introducing a shrinkage force. In this paper, we present an improved geodesic active contour model based on the generalized geodesic active contour with an error term. Experimental results show that the proposed model works successfully for concave objects, it can not only detect object boundaries, but also improve double rings. The convergence rate is faster, the robust is better than the generalized geodesic active contour model.