Describing and Matching 2D Shapes by Their Points of Mutual Symmetry

Arjan Kuijper, Ole Fogh Olsen

Abstract: A novel shape descriptor is introduced. It groups pairs of points that share a geometrical property that is based on their mutual symmetry. The descriptor is visualized as a diagonally symmetric diagram with binary valued regions. This diagram is a fingerprint of global symmetry between pairs of points along the shape. The descriptive power of the method is tested on a well-known shape data base containing several classes of shapes and partially occluded shapes. First tests with simple, elementary matching algorithms show good results.