Modified Distance Transform And Linking Algorithm For Image Skeletonization
Chung C. Lee · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1983
This paper describes a modified distance transform (MDT), which combines the original distance transform (DT) with a new set of selection rules to be defined, and an appropriate linking algorithm to produce a connected skeleton in computing times that are significantly shorter than the medial axis transform (MAT) implementation for large image arrays. The new set of selection rules is applied over an extended neighborhood and makes the skeleton generated more connected, especially for branched images. These selection rules mostly remedy the limitations of the original DT technique but not entirely, and so a linking algorithm is needed. The linking algorithm described operates on a different size of window larger than the DT window for each skeleton element, but the linking process is operated on a smaller subset of the original image area; i.e., the existing skeleton elements indicated by the new selection rules. In order to minimize the processing time for the linking process, a linearity test is introduced that further reduces the application of the linking process to a subset of linear skeleton elements.