A fast compression-based similarity measure with applications to content-based image retrieval