1989 Reports

A regularized solution of shape from shadows

Hatzitheodorou, Michael; Jackowski, Tomasz; Papageorgiou, Anargyros

We present a regularized solution to the shape from shadows problem. In this problem the shadows cast on an unknown surface yield data that can be used for the reconstruction of this surface. In the simulation presented here we assume that the data can now be perturbed by noise. It is shown that the regularized approach produces a solution that can handle noisy information while being very similar to the solution obtained by the approximation theoretic approaches used in earlier work. We provide implementation runs where the performance of the algorithm in recovering unknown surfaces is tested. Furthermore, we study the visual effects of smoothing on the various reconstructions.