基于级联变换的光学和SAR图像配准算法

针对光学和SAR(Synthetic Aperture Radar)图像配准中存在明显辐射和几何差异的问题,提出了一种基于级联变换的多源遥感图像配准方法.首先,利用灰度变换提取光学和SAR图像间的稳定结构特征,去除辐射差异性;然后,提出一种新的加权对数极坐标变换算法,解决图像间全局几何差异性,保证算法的尺度和旋转不变性,并初步得到整体的平移量;最后,通过局部几何变换,得到一系列的匹配点对,构建薄板样条模型,实现图像的精确配准.实验验证了算法去除辐射差异性和获得全局几何变换参数的能力,与传统的多源图像配准算法相比,基于级联变换的配准算法鲁棒性好,配准精度高.

Robust registration method of SAR and optical remote sensing Images based on cascade transforms

Because that there are obvious differences of radiometric and geometric in SAR and optical image registration, this article presents a method based on cascade transform (CT) to register SAR and optical images. To remove radiation differences, gray transformation was used to extract stable structure information for the SAR and optical images. Then, Weighted-Logarithmic-Polar transformation is proposed to obtain scale and rotation parameters between the images, solving the global geometric differences. Finally, local geometric transform is conducted to get accurate matches and the image was registered by TPS model. The registration experiments show that the proposed method is able to extract stable structures and robust to the differences of radiometric and geometric. The precision of the registration method is high.