Hotelling trace criterion as a model for the performance of an IR imaging system

The effect of fixed-pattern noise on the detection of subtle targets embedded in Gaussian white noise was investigated, comparing human detection performance with that of an ideal linear observer characterized by the Hotelling Trace Criterion (HTC). Five signal-to-noise ratios (SNR) were tested for the two conditions: with and without fixed-pattern noise, where both sets of images also contained Gaussian white noise. In this study, the SNR was calculated using the Gaussian white noise only; the additional fixed pattern noise was not figured into the SNR. Human detection performance did not differ significantly in the two conditions. In both conditions performance increased significantly with increases in SNR. The HTC figure of merit predicted accurately the experimental results for the fixed-pattern noise.