A single framework for multispectral image indexing in the wavelet-based compressed domain

In this paper, we are interested in designing a content-based image retrieval system of multichannel images in the wavelet-based compressed domain. The considered images are assumed to be compressed through an appropriate multi-scale decomposition and the salient features are extracted from suitable statistical models of the coefficients. The originality of our approach consists in addressing the problem in a unified framework. Indeed, we study the relevance of coupling the multiscale decomposition to an efficient feature extraction and a retrieval procedure.