Histogram-based object recognition using shape-from-shading

This paper presents an experimental study of the problem of object recognition using surface attributes delivered by shape-from-shading. We focus our attention on histogram-based representations. Several surface attributes are considered. These include slant and tilt angles, mean and Gaussian curvature, and, shape-index and curvedness. Our study reveals that recognition rates of 98% are achievable under controlled lighting conditions. When there is significant illumination variation then recognition rates of 80% are achievable.