一种加权计算基础矩阵的高精度算法

A High-accuracy Algorithm for Weighted Computing Fundamental Matrix

分析了图像匹配点对基础矩阵的不同影响,引入具有明显几何意义的匹配点到对极线的距离作为匹配点的加权因子,将匹配点集合看作随机变量并将加权因子作为匹配点的概率分布。在此基础上对原始图像数据进行归一化处理,利用8点算法得到基础矩阵。大量的试验结果表明,该方法明显降低了计算余差,提高了基础矩阵的计算精度。

This paper analyses the different influence on the fundamental matrix （F matrix） by the different matched points in images and the weighted factor is introduced to express the distance between the matched point and the epipolar line which has the obvious geometry meaning. The paper looks the set of matched points as a random variable and the weighted factor as the probability distribution. Based on this, the normalization is processed to the origin image data and the F matrix is gotten using the 8-point algorithm. Experimental results show that this algorithm obviously deduces the residual errors and the calculated precision of F matrix is improved notably.