Corner Preserved One-Pass Parallel Thinning Algotithm with the Aids of Condensed Image Plane

Abstract

A new one-pass parallel thinning algorithm is proposed in this paper. With the aids of the condensed image plane, the proposed algotithm obtains skeletons with the properties of perfectly 8-connected, noise-insensitive, and topologically equivalent to the original object shape without excessive erosion by performing template matching on 21 newly designed templates. Futhermore, the corners and T-junctions are perfectly preserved in the obtained skeletons. It is helpful for many image analysis and recognition tasks simce both corner and T-junction are important structural features for these tasks. Experimental results show that this algorithm is indeed in good performance.