中文摘要：

全仿射形变条件下,待配准合成孔径雷达（synthetic aperture radar,SAR）图像与参考SAR图像之间存在各向异性尺度变化,导致传统的点特征图像配准算法难以提取到足够多的匹配特征点进行图像配准。为此,提出了一种基于仿射形变矩阵分解与尺度变化矩阵估计的点特征图像配准算法。该方法首先将仿射形变矩阵分解为图像旋转矩阵、尺度变化矩阵以及常数矩阵的乘积,而后利用粒子群优化（particle swarm optimization,PSO）算法对尺度变化矩阵中的未知参数进行搜索估计,并根据估计结果对图像进行尺度规范处理,以抑制图像间的各向异性尺度变化,在此基础上再利用尺度不变特征转换（scale invariant feature transform,SIFT）算子提取匹配特征点进行配准处理。实验结果表明,与现有方法相比,对于全仿射形变条件下的SAR图像配准,本文所述算法可以提取到更多的匹配特征点,因而具有更好的配准性能。

英文摘要：

In the fully affine synthetic aperture radar （SAR） image registration conditions, the scale change between reference images and registering images is non-isotropy, which makes it difficult to extract enough matching feature points for the traditional image registration method based on feature points. To deal with this problem, a new image registration algorithm based on feature points is proposed. The affine matrix is first decomposed into products of image rotation matrixes, scale change matrixes, and constant matrixes. Then the un known parameters in scale change matrixes are estimated by the particle swarm optimization （PSO） method. Based on the estimation result, reference and registering images are normalized to suppress the non-isotropy scale change between them. After that, the scale invariant feature transform （SIFT） operator is employed to extract matching feature points, and the image registration is based on it. The experimental results show that, for the fully affine SAR image registration, the proposed algorithm can obtain more matching feature points than the existed methods, so it has a better performance.