针对基于场景的自适应校正算法普遍存在鬼影的问题, 分析了神经网络算法(NN-NUC)产生鬼影的原因，并在此基础上提出了用基于偏微分方程(PDE)的非线性滤波方法取代NN-NUC算法中邻域平均的方法来获取期望图像,从而减少边缘像素误差，达到消除鬼影的目的.采用实际采集的红外图像进行实验，结果表明,很好地消除了鬼影.与已有的几种去鬼影的方法相比，具有更快的收敛性.

Generally, most of adaptive nonuniformity correction algorithms have the ghosting artifact problem. In this paper, the cause of ghosting artifacts in Neural Network nonuniformity correction (NN-NUC) algorithm for infrared focal plane array (IRFPA) was studied. Based on the analysis, a novel algorithm for eliminating the ghosting artifact was proposed, which replaces the linear spatial average filter in the NN-NUC algorithm with the partial differential equation (PDE)-based nonlinear filter to estimate the desired image. The comparison experiment using real IRFPA infrared image shows that the proposed algorithm can effectively remove the ghosting artifact. Compared with several deghosting algorithms, the proposed algorithm converges much faster.