Some (more) results

Reconstructing a toy squirrel from 32 images (648x480 pixel²) taken under natural lighting condition with cluttered background (or noisy silhouettes): left, one view of the object; right, our reconstruction results. Some comparisons are in the paper.

A result not reported in the paper-reconstructing a skull from 23 calibrated images (data courtesy of Y. Furukawa, UIUC); left, one of the 23 images (~2000x2000 pixel²); middle and right, our and Furukawa et. al.'s reconstructions respectively (shown with different visualization tools).