Quadtree based image indexing in wavelets compressed domain

A considerable amount of research has been done in the past on the use of quadtrees in pixel domain for image indexing as well as their use with the wavelet decomposition for image compression. In this paper we attempt to fuse those two approaches in order to produce a system which makes use of quadtrees for image indexing in wavelets compressed domain. The proposed system uses the quadtree representation of the DWT significance map as the indexing key. The measure of similarity between two images is given by the comparison (XOR-ing) of their respective quadtree structures. Extensive experiments have been made using a database of over 1000 images. The accuracy of the system is demonstrated by a representative set of results as well as a comparison with a state of the art benchmark system.