Typical Segment Descriptors: A New Method for Shape Description and Identification

摘要

In this paper we introduce a new method for recognizing and classifying images based on concepts derived from Logical Combinatorial Pattern Recognition (LCPR). The concept of Typical Segment Descriptor (TSD) is introduced, and algorithms are presented to compute TSDs sets from several chain code representations, like the Freeman chain code, the first differences chain code, and the vertex chain code. The typical segment descriptors of a shape are invariant to changes in the starting point, translations and rotations, and can be used for partial occlusion detection. We show several results of shape description problems pointing out the reduction in the length of the description achieved.