Boosting Associated Pairing Comparison Features for pedestrian detection

This paper proposes a novel approach to boost a set of Associated Pairing Comparison Features (APCFs) in Granular Space for pedestrian detection, in which Pairing Comparison of Color (PCC) and Pairing Comparison of Gradient (PCG) are two kinds of essential elements. A PCC is a Boolean color comparison of two granules and a PCG is a Boolean gradient comparison of two granules, which is motivated by animal vision system that using simple comparison information in both color and gradient modes for visual perception. Unlike previous works that describe object shape, our method is to find the symbiosis of colors or gradient orientations. Experiments on multi-view multi-pose pedestrian data demonstrate the efficacy of the proposed approach.