Development of a Video-Rate Stereo Machine

Abstract

A video-rate stereo machine has been developed at Carnegie Mellon University with the capability of generating a dense range map, aligned with an intensity image, at the video rate. The target performance of the CMU videorate stereo machine is: 1) multi image input of 6 cameras; 2) high throughput of 30 million point×disparity measurment per second; 3) high frame rate of 30 [frame/sec] ; 4) a dense depth map of 240×256; 5) disparity search range of up to 60 [pixel] ; 6) high precision of up to 8 [bits] (with interpolation) ; 7) uncertainty estimation available for each pixel. The stereo machine is based on a multi-baseline stereo method. The basic theory requires some extensions to allow for parallel, low-cost, high-speed machine implementation. The three major ones are: 1) the use of small integers for image data representation: 2) the use of absolute values instead of squares in the SSD computation (SAD instead of SSD) ; and 3) camera geometry compensation capability for precise camera alignment. The stereo machine successfully generates disparity map of 200×200 with 23 steps of disparity search range in the video rate.