Work

BWT-based efficient shape matching

Year: 2007
Type: conference-paper
Abstract: Effective shape-based image retrieval requires an appropriate representation of object shape contours. Such a representation should be invariant under certain transformations, such as those due to rotation, scaling, partial occlusion, noise in the image, or changes in the viewing geometry. Given a shape boundary, we decompose it into primitive shape segments that capture the saliency of object parts, and perform retrieval based on the primitives. Motivated by the sorted contexts of the Burrows-Wheeler Transform, we present an algorithm for efficient shape matching, suitable for large-scale shape databases, when the shape boundaries are represented as a sequence of shape primitives. Given a query shape, the algorithm can locate all the potential areas in the database where a match could occur in time that is logarithmic with respect to the database size. The potential matches are then verified in time that is linear with respect to the number of potential matches. Performance of the proposed algorithm is evaluated using both synthetic and real shape databases.
Language: English
Open Access status: green
OpenAlex ID: W1997243233
FWCI:
Cites: 33
Cited by: 3
Related to: 10