遥感面阵凝视成像系统可以得到同一场景的多幅图像, 研究者常利用这一特点进行多幅图像超分辨重建, 以提高遥感图像空间分辨率。但是这类研究往往将超分辨过程独立出来, 很少结合成像系统的几何参数优化超分辨重建模型。因此, 对成像姿态影响图像不同方向上分辨率的问题进行了分析, 提出了基于姿态角的各向异性模糊估计, 使退化模型更加准确。同时, 为了进一步精确面阵凝视成像系统超分辨重建中的匹配参数估计, 提高由系统引起的全局初始匹配误差的包容性, 基于最大后验法提出并行优化超分辨率图像和匹配参数的方法。算法充分利用成像过程信息并实时优化匹配参数, 实验结果证明与现有方法相比, 不仅可以得到细节信息更丰富, 更易于人眼观察的遥感图像, 并且均方误差降低0.3倍左右, 信息熵平均提高1.2。

A group of images for one scene can be obtained by plane array staring remote sensing system. So researchers often use multi-frame image super-resolution reconstruction to produce images with higher spatial resolution. However, most of reports regard super-resolution reconstruction as an isolated part ignoring that geometric parameters of imaging system have the ability to optimize the reconstruction model. Therefore, the influence of attitude angle on resolution changes in different directions was analyzed particularly. Meanwhile, the corresponding anisotropic blur estimation was presented to improve the accuracy of discrimination model. Because of the matching parameter as a significant role in super-resolution reconstruction, for improving the accuracy of the matching parameter estimation and decreasing the global initial matching error caused by the system, the algorithm of simultaneously optimizing super-resolution image and matching parameters based on maximum a posteriori estimation was proposed. This method takes advantage of the useful information of imaging system and improve the robustness of matching parameter by synchronous optimizing. The experimental results demonstrate that the method of our paper is better than existed algorithms in detail information and definition observing by eyes. In addition, the mean square error was reduced 0.3 times, and the information entropy was increased 1.2 in average.