中文摘要：

针对在图像匹配中，随机抽样一致性（RANSAC）算法对匹配点提纯存在计算量大、效率低的问题，采用将基本矩阵作为模型参数估计对象的方法，对RANSAC匹配点提纯算法进行了改进．在改进的算法中，运用Bucket分割技术抽取粗匹配点对，进行两幅图像的检测角点和粗匹配，利用视差梯度对匹配点样本预检验．实验结果表明，此方法在保证较高精度和鲁棒性的情况下，运算量大幅度减少，提高了图像匹配的速度．

英文摘要：

Improvement has been made for RANSAC alogorithm of matched points purifying in the image matching process by using fundamental matrix as object of model parameter in finding a solution to the problem with the large amount of calculation and the low efficienty that eonerge. By extracting rough matched points in sub-block generate from Bucketing techniques, algorithm test two images with corner detection and rough matching has also been improved. Moreover, it pretests the sample of matching points using disparity constraint. The experiment shows that this algorithm reduces the a- mount of computation largely, improves the speed of image matching and keeps high precision and robust.