针对大幅面图像的拼接问题，提出了一种新的基于特征点的柱面全景图像快速拼接算法。该算法在Harris角检测算法提取特征点的基础之上，针对传统RANSAC（random sampling consensus）算法在提纯匹配点时计算极其复杂，难以实现快速拼接的问题，设计了专门的聚类预筛选的方法进行图像特征点的预匹配，显著提高了特征点匹配的效率；在图像融合部分提出了最佳路径与HSI颜色空间的亮度权重函数相结合的算法，既有效消除了拼接图像中的鬼影现象，又使拼接图像达到了平滑融合效果。实验结果表明，该算法具有匹配精度高、鲁棒性强、计算效率高的特点。

A fast image mosaic algorithm based on feature points matching with excellent results is presented in this paper and two outstanding methods in this algorithm are proposed. The first one is importing a new filtering method for matching points by choosing pairs of correlation feature points with clustering algorithm aiming at the disadvantage of RANSAC algorithm. The efficiency of the algorithm is enhanced by doing this. The other one is raising another new method of blending images by using optimum path best-matched-line combined with pixel brightness weighting function in HSI color space.As a result，the gusting phenomenon is removed and a brightness blending image is achieved. Several results show that this is a highly-matched robust image mosaic algorithm.After choosing data with clustering algorithm beforehand，the efficiency is greatly enhanced.And this algorithm also has prevented ghosting phenomenon and can make images blending naturally at stitching areas.