A method of registration based on skeleton for 2-D shapes

The iterative closest point (ICP) algorithm is an accurate approach for the registration between two point sets on the same scale. However, number and noise of two point sets restrict good performance of ICP algorithm. This paper proposes a novel ICP algorithm based on skeleton (SKICP). The proposed algorithm is to denoise and speed up the point set matching process using skeleton of multi-scale point sets. Firstly, we extract the sparse skeletons from the lower resolution original point set, which have fewer points including its structure features. Secondly, the point set of skeletons is quickly matched in lower resolution, and an initial transformation matrix between two point sets acquired. Finally, the initial transformation matrix is used as the initial value for a more precise registration at high resolution using less iterations. Experiments demonstrate the SKICP algorithm has faster speed and better robustness on 2-D Shapes point set than the traditional ICP algorithm.