Fractal coding has been proved useful for image compression. It is also proved effective for content-based image retrieval. A block-constrained fractal coding scheme and a matching strategy for content-based image retrieval were presented. In the coding scheme, an image is partitioned into non-overlap blocks of a size close to that of a query iconic image. Fractal codes are generated for each block independently. In the similarity measure of fractal codes, an improved nona-tree decomposition scheme is adopted to avoid matching the fractal codes globally in order to reduce computational complexity. The experimental results show that the coding scheme and matching strategy are useful for image retrieval.
|Number of pages
|Tien Tzu Hsueh Pao/Acta Electronica Sinica
|Published - 1 Jun 2000
ASJC Scopus subject areas
- Electrical and Electronic Engineering