论文部分内容阅读
分别分析了红外焦平面阵列(IRFPA)基于定标的非均匀性校正法(NUC)和基于场景的NUC算法各自的优势和问题,在此基础上提出了联合非均匀性校正方法。根据上电时刻焦平面衬底的温度值,从FLASH中提取事先存储的对应温度区间的增益和偏置校正参数,初步消除探测器的非均匀性。通过分析初步校正后图像残余非均匀性噪声的特性,提出了用具有保边缘特性的P-M滤波取代传统神经网络算法中的四邻域均值滤波来获得期望图像,从而减小了图像边缘误差。实验结果表明,该算法收敛速度快,校正精度高,有效避免了因红外焦平面响应特性漂移而引起的图像降质。
The respective advantages and problems of IRFPA calibration based non-uniformity correction (NUC) and scene-based NUC are analyzed respectively. On this basis, a joint nonuniformity correction method is proposed. According to the temperature of the focal plane substrate at the power-on, the pre-stored corresponding temperature range gain and offset correction parameters are extracted from the FLASH to initially eliminate the detector non-uniformity. By analyzing the characteristics of the initial corrected residual nonuniformity noise, we propose to replace the traditional four-neighborhood averaging filter in the traditional neural network algorithm with the P-M filter with guaranteed edge to obtain the desired image and reduce the image edge error. The experimental results show that the proposed algorithm has the advantages of fast convergence and high accuracy, which can effectively avoid the degradation of the image caused by the drift of response characteristic of infrared focal plane.