A micropower centroiding vision processor

Abstract

A biologically-inspired hybrid vision chip is presented for real-time object-based processing for tasks such as centroiding, sizing and counting of enclosed objects. This system presents the first silicon retina capable of centroiding and sizing multiple objects in true parallel fashion. Based on a novel distributed algorithm, this approach uses the input image to enclose a feedback loop to realize a data-driven pulsating action. The sensor provides a resolution of 48 × 48 pixels with a 85 μm × 85 μm pixel footprint and has been measured to consume 243 μW at 1.8-V supply, achieving an equivalent computational efficiency of 724.64 MIPS/mW with a 500-μs process time. © 2006 IEEE.