采用贝叶斯框架的非均匀性校正方法(英文)

A nonuniformity correction method based on Bayesian framework

提出了一种在基于场景非均匀性校正和定标非均匀性校正之间建立桥梁的思路,利用定标法提供的大量先验信息解决收敛速度和鬼影的矛盾问题.利用贝叶斯方法计算非均匀性参数的正确概率,用参数正确概率来决定是否使用该组参数进行校正,从而在源头上抑制鬼影.对于先验概率,定义了非均匀性的局部同分布约束,并通过定标统计的策略利用该约束构建了先验概率;对于观测概率,发现并详细分析了红外焦平面阵列固有的非均匀增益参数空间相关性,利用空间相关性构建了观测概率.最终,通过本文算法对真实和仿真的红外图像序列进行处理,表明该算法在保证高收敛速度前提下,其参数正确概率有效抑制了鬼影,取得了好的处理效果.

In this study, we have created a bridge, which can connect the reference-based NUC and scene-based NUC. The right probability of the scene-based NUC parameters was calculated based on the Bayesian framework. The right probability composed of prior and observation probability was used to determine whether the calculated scene-based NUC parameters are suitable to correct the nonuniformity. The local same distribution constraint is defined in this paper, and the Infrared Focal Plane （IRFPA） gain space relativity has been discovered from the reference-based parameters by this paper firstly. The Bayesian prior probability is mainly determined by the local same distribution constraint, and the Bayesian observation probability is mainly determined by the IRFPA gain space relativity. This method can effectively balance the relationship between convergence speed and ghosting artifacts. Finally, the real and simulated infrared image sequences have been applied to demonstrate our algorithm＇ s positive effect.