11:15 Parallel Dynamic Programming for Realtime 3D Acquisition – S. Yamazaki (AIST)

We propose a method of real-time 3D shape acquisition using a pair of projector and video camera. We encode projector coordinates into a single color pattern using the De Bruijn sequence, and decode it from acquired images robustly by our parallel dynamic programming algorithm. Dense depth map can be reconstructed from each video frame, which allows us to capture the 3D shapes of dynamic objects using commodity hardware. We have implemented our algorithm on NVidia GeForce 280GTX, and demonstrated accurate depth acquisition in VGA resolution at 30Hz.