Share Email Print
cover

Proceedings Paper

Refinement of axial shape description
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

The representation and characterization of planar shapes or regions has important consequences for image processing and image content understanding. Numerous approaches have been developed to provide image analysis with efficient methods to represent shapes. We present a region-based approach to extract a refined axial regional shape description in the form of a skeleton, which is based on the use of the constrained Delaunay triangulation (CDT) and the chordal axis transform (CAT). We elaborate on the exploitation of the approximate edge co-circularity criterion that is used to refine CAT-produced skeletons. The co-circularity criterion enables efficient evaluation of CDT-generated triangle edges lying inside regions (chords), and filters out non-salient chords. The application of this criterion produces smoother skeletons, allows skeleton to have vertices with branching degree higher than three that was due to the use of CDT, and significantly reduces number of skeleton segments. In this paper, in contrast with the chord strength evaluation of the original skeleton rectification algorithm, where chord strength evaluation does not include strengths of its neighboring chords, we introduce smoothing operator to evaluate chord strength by processing chord strengths within local neighborhood. The result of region characterization based on the proposed smoothing-based chord evaluation is that skeleton is more authentic (sensitive) to original shape, while at the same time it preserves all the advantages of the original skeleton rectification scheme. A number of examples, including comparison with skeletons generated by the original CAT-skeleton rectification algorithm, are presented to demonstrate our approach at work.

Paper Details

Date Published: 18 January 2006
PDF: 11 pages
Proc. SPIE 6066, Vision Geometry XIV, 606605 (18 January 2006); doi: 10.1117/12.650822
Show Author Affiliations
Alexei N. Skourikhine, Los Alamos National Lab. (United States)


Published in SPIE Proceedings Vol. 6066:
Vision Geometry XIV
Longin Jan Latecki; David M. Mount; Angela Y. Wu, Editor(s)

© SPIE. Terms of Use
Back to Top