A Method Combining Active Stereo with Global Matching Algorithm for Accurate Depth Extraction
Manqing Lin, Rongke Liu, Yu Chen Pan, Qiuchen Du, Peijuan Chen · 2016
In the field of computer vision, stereo is one of the most popular approaches for depth acquisition. In order to reduce mismatching caused by less-textured scene, active stereo vision has been frequently employed. To the best of our knowledge, active method is usually used in local stereo matching, but it is rarely available in the global matching. As for the stereo, global matching possesses many advantages compared with local matching, such as high quality results, strong resistance of interference and occlusion. This paper proposes a new method for extracting depth images with higher accuracy and better quality by combining active stereo vision with graph-cuts-based global matching algorithm. We demonstrate the theoretical basis of the combination and present the strengths of our method. An improved algorithm based on projection features of active stereo vision for the coefficients of smooth item is also proposed in this paper. We experimentally demonstrate the effectiveness of our approach.