Multiple-shot person re-identification by chromatic and epitomic analyses

► We propose a novel appearance-based method for person re-identification. ► It condenses a set of frames of the same individual into a informative signature. ► It incorporates complementary global and local statistical descriptions of the human appearance. ► Semantic segmentation of objects is exploited to define a part-based descriptor. ► The resulting descriptor define state-of-the-art results on the considered datasets.