TOWARDS CASUAL APPEARANCE CAPTURE BY REFLECTANCE SYMMETRY

Abstract

The visual appearance of a real-world scene or object depends on shape, reflectance, illumination and viewing angle. Given images of a scene, it is possible to reverse the image formation process so as to infer shape and/or reflectance, the key of which lies in an appropriate reflectance model. This thesis exploits several reflectance symmetry properties found in real-world materials to tackle the problem of shape reconstruction and reflectance estimation.