A reconfigurable disparity engine for stereovision in Advanced Driver Assistance Systems

Abstract

Depth extraction in stereovision applications is very time-consuming and requires hardware acceleration in real-time context. A large number of methods have been proposed to handle this task. Each method answers more or less to real-time constraints, depending on the applicative context and user's needs. Thus, flexibility is a strong requirement for a generic hardware acceleration solution, particularly when ASIC implementation is targeted. This paper presents REEFS, a reconfigurable architecture for embedded real-time stereovision applications. This architecture is composed of three reconfigurable modules that enable flexibility at each step of depth extraction, from correlation window size to the matching method. It generates VGA depth maps with 64 disparity levels at almost 87 frames per second, answering hard real-time requirements, like in Advanced Driver Assistance Systems. © 2010 Springer-Verlag.