Skeleton extraction method based on distance transform

The skeleton can describe an object´s geometry and topology with few data, and is applied in a variety of tasks in computer vision. While common distance transform can hardly guarantee the connectivity property of the skeleton, or thinning algorithms for skeleton extraction can´t guarantee the accuracy. In this paper, a new skeleton extraction method is proposed. Based on Euclidean distance transform, the seeds of the skeleton are determined according to the number of greater direction. And then a two-step skeleton growth is employed to obtain connected and one-pixel width skeleton. The experiments prove that the proposed algorithm not only has low time complexity, but also guarantees the connectivity and one-pixel width of the skeleton.