对 SIFT算法进行研究, 针对 SIFT特征描述符的高维数和高复杂度问题, 进行了改进。通过对大量的不同类型的图像进行特征匹配实验, 实验结果表明, 当图像存在不同程度的几何变形、 辐射畸变和噪声影响时, 改进后的算法更稳定、 更快速。-Study of the SIFT algorithm, SIFT feature descr iptor for the high dimension and highly complex issues, has been improved. Through a large number of different types of image feature matching experimental results show that when there are different degrees of geometric image distortion, distortion and noise radiation, the improved algorithm is more stable and fast.