红外技术发展到今天, 红外凝视焦平面探测器阵列性能高、使用简单, 从而成为红外系统的主流。但是, 红外探测器的工艺和技术生产不出像可见光 CCD那样均匀的红外器件, 红外探测器阵列的非均匀性一直是红外凝视探测器的主要缺陷。人们开发了多种非均匀性校正算法, 尤其是基于场景的自适应算法, 极大地弥补了探测器的非均匀性缺陷。但是到目前为止, 各种算法都有一定的局限性, 尚不能彻底解决非均匀性问题。针对目前常用的几种非均匀性校正算法, 包括时域高通滤波算法、神经网络算法、恒定统计量算法等, 在天空、地面等不同场景条件下进行了仿真测试, 对算法的实施效果进行了对比分析。

Nowadays IRFPA(infrared focal plane array)is the main trend in infrared system. Because of technique limitation, IRFPA can’t be as perfect as visible detectors. Nonuniformity has been the main defect of IRFPA for a long time. Many nonuniformity correction algorithms were developed in past decades, especially scene-based NUC algorithm. These algorithms remedy the defect of IRFPA largely. Up to now, every algorithm has its limitation and can’t solve the nonuniformity problem thoroughly. In this paper, several NUC algorithms in common use, include temporal high-pass filter algorithm, neural network algorithm, constant-statistics constrain algorithm and moving scene-based algorithm, are evaluated with sky, cloud, and water surface scenes, and comparative analysis is done with the results of these algorithms.