MRTD: to NUC or not to NUC?

Abstract

We applied a simple method to estimate the Minimum Resolvable Temperature Difference (MRTD) of an LWIR and an MWIR camera. A so-called Siemens star, in our case a thin, black aluminum plate framing a circle that is missing (cut out) every other spoke, is mounted in front of a black body whose temperature is relatively close to room temperature. From short recordings of the black body and Siemens star both the Noise Equivalent Temperature Difference (NETD) and the Modulation Transfer Function (MTF) are extracted and a simple estimate of MRTD = NETD/MTF is obtained. The imaged Siemens star almost completely covers the focal plane array; hence, an MRTD curve for the whole array is obtained. We investigated the effect of Non-Uniformity Correction (NUC) and Bad-Pixel Removal (BPX), two often applied pre-processing techniques, on the MRTD estimate. We find that (1) BPX has only limited effect on the result; (2) NUC is required to obtain a good MTF; and (3) NUC is not a prerequisite to obtain a good NETD estimate, but this is contingent on having a proper segmentation tool or template available. Without a segmentation algorithm, NUC together with simple intensity thresholding provides a sufficiently good segmentation and accordingly a good estimate of NETD.