Frame-rate Robust Stereo on a PCI Board

Many vision applications would benet from frame- rate, dense stereo. In this paper we present a robust algorithm implemented on a single board for a PC. The algorithm used is census transform correlation, which is well-suited for hardware implementation. Un- like existing frame-rate stereo methods, census trans- form correlation is robust to outliers, and does not as- sume that corresponding points have constant bright- ness. We provide an empirical comparison between our method and other frame-rate stereo algorithms us- ing real data with ground truth. The stereo compu- tation is performed on a recongurable computer built from programmable logic on a standard PCI card. The resulting system computes 32 stereo disparities on 320 by 240 pixel images at 46 frames per second. This is the highest performance stereo vision system described to date.