Test of a model of foveal vision by using simulations.

The appearance of four different images from three different distances was simulated by using the individual contrast sensitivity functions (CSF's) of normally sighted observers. The simulations were generated by using the observers' CSF's as a threshold in a pyramidal vision model of band-limited local contrast [J. Opt. Soc. Am A 7, 2030 (1990)]. Simulations based on CSF's obtained in an orientation discrimination task underestimated the observer's sensitivity in discriminating the images. Simulations based on CSF's obtained in a detection task provided a good estimate of observer's performance. The testing method was shown to be sensitive enough to be affected by the high-frequency residual, which is frequently ignored in visual models and simulations. An image-dependence effect found when the high-frequency residual was present was eliminated when the residual artifact was removed. The simulations based on the pyramidal vision model accurately predicted the distance at which they were discriminated from the original image, and thus this model may also serve as the basis for image-quality metrics. The testing method developed can also be used to determine the type of CSF that best represents observer performance in a task.