Geometrical active contour model based on structure tensor

An improved geometrical active contour (GAC) model based on the property of local image structure information contained by scatter matrix is presented. The edge stopping function, which inhibits the evolution at the edges of corners in an image, is composed of the measure of the local coherence H and the eigenvalue. The experiments presented on synthetic and real images demonstrate that the proposed model is more robust against the corners of the boundaries and ensures that the evolving curve of the structure tensor energy model reaches more rapidly the true edge boundaries of the objects after comparing the evolving iterations.