Author(s): John Rasku and Matthew O. Ward, Computer Science Department, Worcester Polytechnic Institute, Worcester, MA 01609

Source: Proc. of DICTA '93, Sydney, Australia, December, 1993.

Abstract: Approximate shape comparison involves comparing two shapes which differ in some aspects, but share some similarities. Typical regions of shape difference result from image noise, occlusion, deformations, and affine transformations. This research presents a visual approach for relating shape difference using Correlation Images. A Correlation Image (CI) is a powerful mechanism for visualizing similarities and differences in ordered data by highlighting sequence data relationships using easily perceptible geometric shapes. The capabilities of four shape sequence representations, extracted from 2D contours of shapes and variants generated by various deformation processes, are explored for relating shape change using Correlation Images. In addition, a method to extend the technique to 3D shapes is presented.