本文提出一种全景图像自动拼接算法，能够实现特征点对的精确筛选和匹配，以及输入图像的自动排序和拼接。首先提取输入图像的尺度不变特征变换SIFT特征点，并采用kd树搜索得到图像之间所有初始特征匹配对；利用欧氏距离比值和中值滤波器对初始特征点对进行筛选后，再应用随机抽样一致算法RANSAC得到图像间精确匹配的特征点对；计算出图像之间的单应性矩阵，在此基础上完成对输入图像的自动排序和配准，最终拼接合成全景图像。实验结果表明，该算法能获得比结合欧氏距离比值的RANSAC算法更高的配准精度，全景图拼接效果较好，具有良好的鲁棒性。

This paper proposes an automatic panoramic images mosaic algorithm, which achieves precise feature filtering and feature matching, as well as the autosorting and autostitching of  the input images. Firstly this algorithm detects the SIFT(Scale Invariant Feature Transform) feature points of the input images, and searches all the initial features of the images to be matched by the Kd tree, after filtering the initial feature points by the Euclidean distance ratio and the Median Filter algorithm, gets the precise feature points with the RANSAC(RANdom SAmple Consensus) algorithm; and calculates the single relation matrix between the images, on the basis of this, completes the image sorting and image registration, then achieves the panoramic images mosaic. The experimental results show that compared to RANSAC with the Euclidean distance ratio, the algorithm can have higher accuracy, better panorama images mosaic and good robustness.