A skeleton extraction framework based on inner-product and border gap

Fibers' Physical properties are influenced by cross-sectional shape. In this paper, a medial-axis based skeleton calculation framework is proposed for cross-sectional shape analysis. In the proposal, border elements of each fiber are numbered sequentially by spatial position; the border gap of each inner pixel is defined to be the length of associated border segment; an 8-connected medial-axis hierarchy is derived with the border gap; branches with minor importance are truncated by vector divergence V d and length-width ratio LWR with SVM classifier. The derived skeleton has good connectivity for various cross-sectional shapes.