Region-Based Object Recognition Using Shape-from-Shading

Abstract

This paper investigates whether regions of uniform surface topography can be extracted from intensity images using shape-from-shading and subsequently used for the purposes of 3D object recognition. We draw on the constant shape-index maximal patch representation of Dorai and Jain. We commence by showing that the resulting shape-index regions are stable under different viewing angles. Based on this observation we investigate the effectiveness of various structural representations and region attributes for 3D object recognition. We show that region curvedness and a string ordering of the regions according to size provides recognition accuracy of about 96%. By polling various recognition schemes, including a graph matching method, we show that a recognition rate of 98–99% is achievable.