Computer Vision Talks at Columbia University

Perception and 3D Reconstruction of Specular Surfaces

Silvio Savarese

Caltech, USA

Interschool Lab, 7th Floor CEPSR 

Host: Prof. Shree Nayar 

Abstract 

One of the main tasks of a visual system is computing the shape of objects. A number of cues, notably stereoscopic disparity, texture gradient, motion parallax, contours and shading, have been shown to carry valuable information on surface shape, and have been studied extensively. Unfortunately, many objects of interest and most man-made surfaces, such as a silver plate, a metal spoon or a clean automobile, are smooth and shiny, violating the hypotheses that underlie the analysis of those cues. However, for specular objects, one additional cue may be precisely the reflection of the enviroment: a deformed picture of the surrounding scene can be seen on the surface of the specular object and amount and type of deformation depend upon its shape. Our research is aimed at understanding how the human visual system uses this cue in the perception of shape and what is the relationship relating deformations and local geometry of the surface. To this effect, we assume a simple calibrated scene composed of lines passing through a point. Under these hypotheses, we demonstrated that local information about the geometry of the surface can be fully recovered up to first order from the orientation of reflected images of at least three intersecting lines. Second and third order information may be recovered from the orientation and curvatures of reflected images of three or four intersecting lines up to one unknown parameter only.