一种用于立体图像匹配的改进稀疏匹配算法

立体匹配有着广泛的应用前景，是计算机视觉领域的研究热点。立体匹配是立体视觉中最为关键和困难的一步，它的目标是计算标识匹配像素位置的视差图。文中提出的立体匹配算法基于置信传播（Belief Propagation，BP）。左图像首先经过非均匀采样，得到一个内容自适应的网格近似表示。算法的关键是使用基于置信传播的立体匹配算法，匹配稀疏的左图像和右图像得到稀疏视差图。通过左图像得到网格，稀疏视差图可以经过简单的插值得到稠密视差图。实验结果表明，该方法与现有稀疏立体匹配技术相比在视差图质量上平均有40％的提高。

An Improved Sparse Matching Algorithm for Stereo Matching

Stereo matching with a wide range of applications is an important research field in computer vision. Stereo matching is also the key and the most difficult problem in stereo vision, its objective is to calculate the disparity map of identified pixel. The proposed stereo matching algorithm is based on belief propagation （BP）. Firstly, a content adaptive mesh is obtained by the non-uniform sampling of the left image. The key issue in the proposed method is to formulate BP, matching the sparse left image and dense right images to get sparse disparity map. We can recover the dense depth map form sparse one due to a simple proposed interpolation method that benefits from the mesh approximation of the left image. The results obtained show that the sparse stereo matching with the existing technology in the quality of depth maps had an average 40% improvement.