中文摘要：

为了提高合成孔径雷达（SAR）图像匹配的稳定性和快速性，提出一种改进的SIFT算法。首先使用Canny边缘检测算法代替原算法提取特征点，对SIFT特征向量主方向进行优化，针对SAR图像的相干斑噪声问题，使用一种阈值自适应的Contourlet变换进行噪声抑制，再使用改进的主成分分析算法对提取的特征向量进行降维。与现有方法相比，不但保持了特征的丰富性，而且简化了计算量，提高了运算速度。实验结果表明：改进的SIFT算法能够在SAR图像发生缩放、旋转等变化的情况下进行准确匹配，具有更快的速度和更高的匹配精度。

英文摘要：

In order to improve the stability and rapidity of synthetic aperture radar（SAR） images matching, an improved SIFT algorithm is proposed. At first, using the Canny edge detection algorithm instead of the original algorithm extracts feature points to optimize the main direction of the SIFI＂ feature vectors, for the problem of speckle noise of SAR images, using a threshold adaptive Contourlet transform to suppress noise and then using the improved principal component analysis（PCA） algorithm to reduce dimensions of features extracted before. Compared with the existing method, the improved SIFT algorithm not only maintains the richness of characteristics but also greatly simplifies the computational complexity and improvs the speed of operation. Experiments shows that the improved SIFT algorithm is able to overcome variations of the scale, rotation and resolution to match two images with faster speed and higher accuracy.