要：图像匹配是计算机视觉中许多领域的基础，特征提取则是图像匹配的基础，其中不变量特征是一

个重要的理论。SIFt是最有效的尺度、旋转、亮度不变量局部特征之一，但算法复杂、计算时间长。分析

了SIFt的计算时间分配，通过计算关键点的邻域梯度直方图时动态修改采样步长，大大提高了SWr的

计算速度。分析了基于SIFt特征的图像匹配算法，提出了双向匹配算法，提高了图像匹配的准确率。实

验结果表明所提出的方法是有效的-Scale invariant feature transform(SIb-T)is one of the most efective local feature of scale，rotation and illumi—

nation invariant，but its algorithm is complicated and computation time is long．The computation time distribution is ana-

lyzed，and the computation speed is greatly improved by dynamically modifying sampling step when computing the gradi—

ent histogram of the region around the key point location．SIFr-based image matching algorithm is analyzed，and a bidi—

rectional matching algorithm is propesed to improve the accuracy of ima ge ma tching．The experimental results show that

the prope sed algorithms are efective．