A disparity refinement using superpixel segmentation in stereo videos

Hui-Yu Huang ； Zhe-Hao Liu

《NCS 2019 全國計算機會議》 2019 (2019/11) Pp. 613-616

https://doi.org/10.6927/NCS.201911.0119

摘要

Stereo matching method is a key technology to make stereo vision in computer vision, such as 3-dimesnion television (3DTV). The disparity obtained in stereo matching can represent the depth information in 3D world and shows the importance in stereo field. Stereo matching estimates disparities from a given stereo image pair. Most global stereo matching methods are computationally expensive and involve many parameters. To compute the cost of the matching points from the sequence of images in stereo matching technology, a local stereo matching is typically used and easy to implement. However, the estimated disparity sequences may cause flickering errors. These errors incur a low visible quality of the synthesized video. In order to solve this problem, in this paper, we propose a segmentation-based disparity refinement method using a local stereo matching technology. The quantitative and qualitative evaluations show that our proposed method can obtain a better result than other methods.

關鍵字

disparity refinement ； segmentation ； stereo matching