Comprehensive wavelet-based image characterization for Content-Based Image Retrieval

A novel image characterization based on the wavelet transform is presented in this paper. Previous works on wavelet-based image characterization have focused on adapting a wavelet basis to an image or an image dataset. We propose in this paper to take one step further: images are characterized with all possible wavelet bases, with a given support. A simple image signature based on the standardized moments of the wavelet coefficient distributions is proposed. This signature can be computed for each possible wavelet filter fast. An image signature map is thus obtained. We propose to use this signature map as an image characterization for Content-Based Image Retrieval (CBIR). High retrieval performance was achieved on a medical, a face detection and a texture dataset: a precision at five of 62.5%, 97.8% and 64.0% was obtained for these datasets, respectively.