最大后验概率算法在遥感图像复原中的应用

Application of Maximum a Posteriori Algorithm in Remote Sensing Image Reconstruction

将最大后验概率（MAP）算法应用于遥感图像复原中。针对算法中系统模糊参数预先不知需要人为设定的不足，通过基于图像微分自相关方法获取点扩散函数，以获取的点扩展函数作为先验知识，分别采用Wiener滤波法、正则MAP法、基于Poisson-Markov模型的MAP（MPMAP）法对遥感图像进行复原，并进行对比。实验结果表明，MPMAP方法复原能力较优，可有效提高遥感图像像质，增强图像的高频部分。

The maximum a posteriori（MAP） algorithm is applied in remote sensing image reconstruction. Considering the fault of MAP algorithm that system fuzzy parameters cannot be known in advance and need to set, autocorrelation of derivative image is used to get the point-spread-function of the system as the prior knowledge. Comparison is carried out among the Wiener filter Image restoration algorithm, regularized MAP image restoration algorithm and MAP image restoration algorithm based on Poisson-Markov model （MPMAP）. The expermental result indicats that MPMAP algorithm restoration ability is the best, which can remarkably improve the remote-sensing image quality and enhance the high frequency part.