SHAPE SIMILARITY MATCHING FOR QUERY-BY-EXAMPLE

This paper describes a unified approach for two-dimensional (2-D) shap e matching and similarity ranking of objects by means of a modal repre sentation. In particular, we propose a new shape-similarity metric in the eigenshape space for object/image retrieval from a visual database via query-by-example. This differs from prior work which performed po int correspondence determination and similarity ranking of shapes in s eparate steps. The proposed method employs selected boundary and/or co ntour points of an object as a coarse-to-fine shape representation, an d does not require extraction of connected boundaries or silhouettes. It is rotation-, translation- and scale-invariant, and can handle mild deformations of objects (e.g. due to partial occlusions or pose varia tions). Results comparing the unified method with an earlier two-step approach using B-spline-based modal matching and Hausdorff distance ra nking are presented on retail and museum catalog style still-image dat abases. (C) 1998 Pattern Recognition Society. Published by Elsevier Sc ience Ltd. All rights reserved.