A novel fast and reliable thinning algorithm

In this paper a new algorithm is proposed for skeletonizing binary digital images. The algorithm does not employ the conventional pixelor non-pixel-based techniques. Instead, it identifies regions of particular shapes in the image and substitutes apppropriate skeleton patterns for them. Initially, as many horizontal and vertical strips as possible are detected. These correspond to rather straight, long and narrow regions in the original image. Any remaining regions correspond to joints between strips. The strips are then grouped into trapezoidal regions which may be replaced by the appropriate skeleton patterns. The individual skeleton patterns are then merged using the skeletons due to the jointing regions, generating the final skeleton. A compact representation is obtained for post-processing, and the issue of robustness with respect to noise is addressed.