A Generalized Computer Vision Chip

✍ Scribed by Richard A. Lane; Neil A. Thacker; Luke Seed; Peter A. Ivey

• Publisher: Elsevier Science
• Year: 1996
• Tongue: English
• Weight: 459 KB
• Volume: 2
• Category: Article
• ISSN: 1077-2014
• DOI: 10.1006/rtim.1996.0021

✦ Synopsis

his paper describes a generalized video convolution processor (VCP) which fulfils the requirements of an algorithm for depth estimation from stereo image pairs allowing the acceleration of Tt his process to near-real-time rates. In particular, the VCP is designed to perform high-precision image rectification. The very high level of mathematical integrity which is required for this operation is not supported by commercial image warping devices. Additionally, the generality of the processor allows it to be used across a wider range of computer vision and image processing applications. A prototype has been fabricated using a two-layer metal 1 µm CMOS process and the chip operates at 30 MHz. The VCP can rectify a 256 ϫ 256 image at 25 frames/sec and allows truly arbitrary image warps.