Real-time recognition of planar shapes is usually based on either invariant shape features or contour matching. Contour matching is a precise but time-consuming operation. In this letter we present a simple but efficient multiresolution technique that considerably speeds up the process by using shape pyramids as an aid to matching. The proposed algorithm is tested with a large set of test shapes. Finally, the algorithm is compared to a composite shape recognition technique that combines a feature based pre-selector and a conventional contour matching algorithm.
- Shape matching
ASJC Scopus subject areas
- Signal Processing
- Computer Vision and Pattern Recognition
- Artificial Intelligence