Recovering shapes of specular objects in motion via normal vector map consistency
Alexey Pak · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2012
We present a novel method to reconstruct the shape of a moving specular object from multiple camera images taken near a calibrated display with a known color-encoded pattern. The method is based on voxel carving exploiting the consistency of the cones of allowed normal vectors, or normal vector maps (NVMs). The algorithm is highly parallel, does not rely on pattern recognition and applies to surfaces of arbitrary complexity.