Abstract: Forsolving the poor connectivity problems of skeleton extraction method based on distance transform, ant colony search skeleton extraction method based on distance transform is proposed. The objects potential skeleton points are obtained according to the distance transform and its gradient.By controlling the polygon approximation errorofthe object boundary curve, the boundary vertices with different contribution to the object shape information are determined,and the convex vertices are takenas the endpoints of the object'sskeletonline.The endpoints of redundant skeleton branches are eliminated according to the inner angle size of the endpoint .The heuristic guidance function is determined by distance transformation gradient, pixel similarity and direction of the skeleton line synthetically, the value of the pathcostis proposed to evaluate the search path. The experiment results show that the method not only has good connectivity and multi-scale conditions controllability, but also reduces the redundant skeleton branches and overcomes the effects of boundary noise at the same time.

Key words: ant colony algorithm, distance transform, gradient, polygonal approximation, skeleton extraction, skeleton pruning