基于反馈的立体匹配算法

Stereo Matching Algorithm Based on Feedback

针对因传统的置信度传播(BP:Belief Propagation)算法生成的深度图并非十分精确而导致深度图的绘制及立体重现存在的失真问题,提出一种基于反馈的立体匹配算法,使生成的深度图进一步优化。该算法首先通过BP算法生成深度图,然后由左视点图像与深度数据绘制右视点图像,将绘制出的右视点图像与真实右视点图像进行比较,再将比较结果反馈给BP算法的匹配代价,最后由新的匹配代价重新生成深度图。实验结果表明,该算法与无反馈的算法比较,绘制右视图的PSNR(Peak Signal to Noise Ration)值平均增高0.2 dB。

As the depth map created by traditional belief propagation（BP：Belief Propagation） algorithm is not very accurate causing the distortion of the depth map rendering and three-dimensional reconstruction.We proposed a stereo matching algorithm based on feedback to optimize the depth map.The traditional BP algorithm is used to get the depth map.We utilize the left eye viewpoint and depth information to render the right eye viewpoint which is to be compared with the real right eye viewpoint and send the result of comparison back to the data cost.The depth map will be created again with the new data cost.Experimental results show that by using the proposed algorithm,the PSNR（Peak Signal to Noise Ration） of the rendered viewpoint will enhance 0.2 dB.