基于RLS的红外焦平面器件非均匀性校正算法

RLS-based Non-uniformity Correction Algorithm of IRFPA

针对红外焦平面器件(IRFPA)存在着严重的非均匀性和探测元响应特性随时间漂移的实际情况,开发了基于场景的递归最小二乘(RLS)非均匀性校正算法,通过逐帧估计每个探测器的增益和偏置,来补偿图像中的固定图案噪声。通过对参考信号的精确估计,达到最优的校正精度。在算法中,首先提出了一种条纹噪声估计方法来消除图像中的条纹噪声;然后利用周围探测单元的像素值,采用自适应加权平均滤波器的方法精确估计目标边缘处的参考信号;权系数是根据加性误差来选择的。这样,随着递归次数的增加,参考信号能够更接近真实的辐射信号。通过参考信号可以精确地估计探测器的增益和偏置。仿真实验以及对实际红外图像序列的实验结果表明,本文提出的校正算法收敛速度快、校正精度高,具有很好的外场工程适应性能。

Aiming at the facts that detector element photo-response characteristic drifts with time and infrared focal plane array （IRFPA） has serious non-uniformity, a new scene-based recursive least square （RLS） non-uniformity correction algorithm for infrared image sequence was developed. The algorithm estimates the gain and the offset of each detector element and compensates for fixed pattern noise （FPN） in a frame-by-frame basis. The key that the algorithm achieves optimum correction quality is the accurate estimation of the reference signal. In order to estimate the reference signal accurately on the edge of object in every frame, an adaptive weighted average filter is proposed. In the average filter, the data applied for averaging come from neighbor detector element correction output and the weights are selected according to additive error deviation. In doing so, the reference signal can approach the real irradiance signal along with the recursive number increase. At the same time, in the paper a stripe noise estimation method is also p IRFPA infrared data show that th roposed using scene data. The comparison experiments for simulated data and real e correction algorithm features fast convergence and high correction precision. The algorithm has good adaptability for field engineering applications.