一种红外探测器频域参考像元校正的设计方法

A Method of Reference Pixel Correction in Fourier Domain for the Infrared Detector

国际上两种较为先进的红外探测器参考像元校正方法——EUCLID中的均值校正法和JWST中的IRS2频域校正法,均无法同时降低白噪声分量与相关噪声分量,为此文章提出了一种新的红外探测器频域参考像元校正方法,以尽可能降低系统噪声。该方法延续频域校正的设计流程,基于系统噪声特性设计出合理的频域系数,再反变换至时域便于工程实现。相比原方法额外加入了谱估计、谱线合并、白噪声滤波等步骤。经MATLAB仿真分析验证,使用新方法可以有效解决IRS2频域校正会放大白噪声分量的问题。在以白噪声为主或白噪声分量与相关噪声分量占比相近的条件下,该方法都可取得比均值校正和IRS2方法更好的降噪效果;而在以相关噪声为主时,校正效果略差于原方法,可以在设计过程中跳过白噪声滤波步骤,使得此方法降噪效果与IRS2法相近。

Two advanced reference pixel correction methods,mean correction method and IRS~2 correction method in Fourier domain,have been proposed internationally.Because these two methods cannot reduce the white noise component and correlated noise component at the same time,a new reference pixel correction in frequency domain method for the infrared detector is proposed to reduce system noise as much as possible.This method continues the design process of frequency domain correction,designs reasonable frequency domain coefficients based on the system noise characteristics,and then reverses them to the time domain,which is convenient for engineering implementation.Compared with the original method,spectral estimation,spectral line combination and white noise filtering are added.The MATLAB simulation results show that this method can effectively solve the problem of white noise component amplification in IRS2 method.Under the condition that white noise is dominant or close to correlation noise component in proportion,this method can achieve better noise reduction effect than the mean correction and IRS2 methods.When correlation noise is dominant,the white noise filtering step can be skipped in the design process to have similar noise reduction effect to the IRS2 method.