提出了一种基于场景的红外图像非均匀性校正算法。该算法结合了两点定标校正算法和基于场景的改进的恒定统计算法,将两点校正算法的校正系数作为恒定统计算法的系数初值,并引入阈值进行运动状态检测,对运动场景和非运动场景分别进行系数更新。实验表明,该算法可以实现对红外图像非均匀性的校正,对于本文实验中的视频图像,在100帧时算法收敛,其收敛时间优于其他传统基于场景的非均匀性校正算法,并一定程度上抑制了“鬼影”现象。

A scene-based non-uniformity correction algorithm for infrared image is introduced in this paper. This algorithm combines two-point correction with scene-based improved constant statistic algorithm. It makes the correction coefficient in two-point correction as the initial value of correction coefficient in improved constant statistic algorithm. By introducing a threshold value to judge the movement of scene, the coefficient can be updated separately for motional scene and motionless scene. The test result shows that this algorithm can achieve the non-uniformity correction for infrared image. The algorithm converges in the looth frame of the test video. The convergence time is shorter than traditional algorithms, and the “ghost” can be suppressed to some extent.