A scalable retrieval of multicomponent images in the wavelet transform domain

This work presents a new method for image retrieval in the compressed domain. For this purpose, the image is firstly decomposed according to a lifting scheme, which provides an exact and compact multiresolution representation of the input image. Then, the parameters of a statistical model of the wavelet coefficient distribution are computed and are considered as the salient features of the corresponding images. Our contribution consists in organizing the search procedure of the best matching index in such a way that the multiresolution representation is fully exploited. Experimental results indicate that the search time has been reduced dramatically by the proposed method.