A wavelet-based approach to fast similar shape retrieval

摘要

This paper presents an efficient wavelet-based approach to similar-shape retrieval. Three primary issues in image retrieval are addiressed: feature extraction and representation, similarity measure and searching methods the proposed algorithms consist of the detection of image feature points in a hierarchical fashion via wavelet tranform, parallel invariant moments calculation and B-spline curve representation. in addition, we introduced a two-stage coarse-to-fine matching scheme to search for the best similar shape represented by its boundary feature points Furthermore, we implemented the proposed algorithms in parallel on a network of workstation clusters so as to tackle the relevant computationally expensive tasks with robustness, reliability and efficiency.