With regard to application-specific constraints such as size, power consumption, and price, many real time vision tasks are still too demanding for today's state-of-the-art hardware. This is especially true for purely digital hardware systems. As an interesting alternative, we investigated the implementation of smart vision algorithms by an application-specific vision chip featuring combined analog and digital processing. The analog part implements computationally intensive operations in a massively parallel array, which allows to realize the remaining operations by a reduced complexity dedicated digital processor. This paper reports on the systematic design and VLSI implementation of dedicated vision chips for automotive image processing, eye tracking, and visual inspection
Date of Conference: 1999