We propose a scale-based method for extracting corners from a given polygonal contour figure. A crystalline flow is introduced to represent geometric features in a scale-space. It is an extension of a usual curvature flow. A special class of polygonal contours is evolved based on the nonlocal curvature. The nonlocal curvature is determined for each facet by its length. In the crystalline flow, a given polygon remains polygonal through the evolving process. Different from a usual curvature flow, it is easy to track a facet in a given polygon through the evolution. This aspect helps us to extract a set of dominant corners. Experimental results show that our method extracts a set of dominant corner facets successfully from a given contour figure.
|ジャーナル||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|出版ステータス||Published - 2003 12月 1|
ASJC Scopus subject areas
- コンピュータ サイエンス（全般）