Finding people using scale, rotation and articulation invariant matching

Abstract

A scale, rotation and articulation invariant method is proposed to match human subjects in images. Different from the widely used pictorial structure scheme, the proposed method directly matches body parts to image regions which are obtained from object independent proposals and successively merged superpixels. Body part region matching is formulated as a graph matching problem. We globally assign a body part candidate to each node on the model graph so that the overall configuration satisfies the spatial layout of a human body plan, part regions have small overlap, and the part coverage follows proper area ratios. The proposed graph model is non-tree and contains high order hyper-edges. We propose an efficient method that finds global optimal solution to the matching problem with a sequence of branch and bound procedures. The experiments show that the proposed method is able to handle arbitrary scale, rotation, articulation and match human subjects in cluttered images. © 2012 Springer-Verlag.