After establishing a working-base image of given pixel-resolution and digitization, in conjunction with a calibration transform, the present method for image-processing based reconnaissance system performance evaluation gives attention to scaling, atmospheric, and MTF factors. The 'kernel spread-point definition', which is a weighted map of the redistribution of energy from a point object, is used to emulate lens-diffraction limitations, linear image motion, and simusoidal or random vibration. Recognition and identification criteria can be evaluated by simply looking at targets.
Gary L. Conrad,
"Reconnaissance system performance prediction through image processing", Proc. SPIE 1342, Airborne Reconnaissance XIV, (1 November 1990); doi: 10.1117/12.23139; https://doi.org/10.1117/12.23139