A Novel Sub-Iterative Parallel Skeletonization Method

A Novel Sub-Iterative Parallel Skeletonization Method

Parallel thinning methods are digital skeletonization approaches that apply parallel strategies to accelerate the processing speeds of algorithms. Existing parallel thinning methods fail to produce clean and complete single-pixel-width skeletons, where clean means that a skeleton contains fewer unwanted branches caused by boundary noise and complete means that the skeleton should have the same topology as the original image. To over-come this problem, in this paper, a novel sub-iterative parallel thinning method is proposed based on the Zhang-Suen (ZS) method by altering the original partial conditions and adding several additional deletion templates and one restoration template to each sub-iteration. Three experiments are conducted to evaluate the performance of the proposed method. The simple pattern experiment shows that the skeleton resulting from the proposed method can maintain the complete original topology. The noise experiment shows that the proposed algorithm is insensitive to boundary noise. Thus, it can produce a relatively clean skeleton. The complicated image experiment shows that the proposed method has a higher thinning rate than other approaches and has application potential in natural images.