立体视觉中误匹配滤波方法的研究

从二维空间和三维空间2种角度研究误匹配滤波算法,提出在匹配前用于降低误匹配的灰度预处理算法和一种基于真实控制点的视差滤波算法。前者只针对2幅图像的重叠区域进行灰度均衡,可以减少计算量,后者在传统视差均值滤波的基础上可进一步提高误匹配的滤波效率。基于真实图像的实验结果表明,新算法可以有效滤除误匹配,提高三维重建精度,保证重建效果。

This paper discusses some filtering algorithms for false matches on the basis of 2D space and 3D space,and puts forward a grey pre-processing algorithm applying beforehand to lower the false matching rate and a disparity filtering algorithm based on ground truth control points.The former balances the overlap area between two images to reduce the calculation quantity,and the latter based on disparity mean value filtering can improve the filtering efficiency. Experimental results of real images prove that the false matches can be eliminated effectively with the new algorithms,ensuring the effect and accuracy of 3D reconstruction.