A Measure for Evaluating Detectability of Pictorial Patterns under Shot-Noise Limited Conditions

Abstract

A novel method to evaluate the quality of pictures is presented. In image pickup systems, including human vision, shot noise is inevitably added to the observed image during the photo-electric conversion process. To quantitatively express the influence of shot noise on the image signal, we propose a new measure of “detectability” of the pattern, based on Iijima's picture observation theory : a logarythmic function of the ratio of the NORM of the pattern to the fluctuation. This detectability reveals that the accuracy limit of blur detection corresponds to the residual focusing error in human visual systems. Also, the measure shows characteristics similar to sharpness, except for extremely short viewing distances.