Insights into local stereo matching: Evaluation of disparity refinement approaches

Local stereo matching as a main category of stereo methods has been studied for decades. Recently, there appeared some well-performed local stereo methods and these state-of-the art methods show comparable performance in terms of quality and better performance in terms of speed with global methods on the Middlebury datasets. However, previous surveys of stereo matching do not consider disparity refinement step, which plays a significant role in improving depth quality for local methods. Thus, in this paper, we focus on the evaluation of the existing disparity refinement approaches and provide some insights into local stereo matching.