Efficient Stereo Matching by Local Linear Filtering and Improved Dynamic Programming.

In this paper, we present a dense stereo correspondence algorithm combining local linear filtering and improved dynamic programming (DP) algorithm, which maintains good performance in both accuracy and speed. Traditional DP, as we all know having most advantage in efficiency among global stereo approaches, suffers from typical streaking artifacts. Recently, an enhanced DP-based algorithm can reduce streak effects well by employing vertical consistency constraint between the scanlines, but with ambiguous matching at object boundaries. To tackle this problem, a cost-filtering framework is deployed as a very fast edge preserving filter in the improved DP optimization, without additional burden of computational complexity. The performance evaluation using the Middlebury benchmark datasets demonstrate that our method produces results comparable to those state-of-the-art algorithms but is much more efficient.