Contour Tracking Using Centroid Distance Signature and Dynamic Programming Method

In this paper, we propose a novel approach based on centroid distance signature and dynamic programming (DP), which can be used to guide a contour tracking execution. Edges are detected by an improved canny method; however, they still contain many noises as well as interferential information. So an edge-based contour tracking method is proposed to link the optimal object´s outline. First, we apply centroid distance function to the obtained edges to get the rough signature graph. Second, the optimal path ofthe signature graph is acquired based on DP algorithm. Last, we resume the optimal contour using the acquired signature trace after inverse transform. The superiority of the proposed model comes from 1) the accurate acquisition of object´s outline based on DP algorithm, and 2) scale and rotational invariant expression of contours with the centroid distance function.