Fast correlation-based stereo matching with the reduction of systematic errors

Abstract

The purpose of this work is to develop a stereo matching algorithm that produces a dense disparity map for a real time vision system. Generally, correlation-based correspondence algorithms produce systematic errors at depth discontinuities, including half occluded areas, and require significant computational power. To solve these problems, we identify the properties of these areas containing matching errors and carry out an error compensation process with an effective implementation.