A distance map based skeletonization algorithm and its application in fiber recognition

The conversion of two-dimensional objects into a skeletal representation is a fundamental calculation in image processing and pattern recognition, because topological structures of objects can be preserved in the skeleton. For the purpose of shaped fiber recognition, the number of shaped fiberpsilas branches is necessary to be calculated. In this paper, a skeletonization algorithm is proposed, in which the fiberpsilas topology information is well preserved. The algorithm also shows good noise tolerance. At the end of this paper, the experimental results of shaped fiber recognition based on proposed algorithm are given.