An Automatic Pruning Method for Skeleton Images

Abstract

Skeletons can be regarded as a compact shape representation in that each pattern can be completely reconstructed from its skeleton. One of the limitations of the application of the skeleton for object analysis and recognition is the existence of redundant skeleton branches. Skeleton pruning is an effective way to remove the redundant skeleton branches in the skeleton images. However, most of the existing pruning methods require manual tuning of the parameter to control the power of pruning, which is not convenient to use. In this paper, we have proposed a fully automatic pruning method that adjusts the power of the pruning according to the image and achieves good pruning results in the experiment.