A Predictor of Human Visual Performance (PREVIP) at Form Discrimination Tasks

This technical report presents an algorithm which implements the form Shape discrimination task performed by a model of the human visual system. This task is important because there is no known method by which the suitability of a symbol set can be evaluated without conducting expensive and time consuming psychophysical tests. The algorithm reported herein is a PREdictor of VIsual Performance PREVIP at visual discrimination tasks. PREVIP is also used to display the results of the spatial frequency filtering performed by the model and to test the modeling of new tasks. PREVIP accepts a digital representation of a set of visual stimuli and applies a preprogrammed visual response filter. PREVIP then uses these filter outputs to compare the stimuli against a set of previously defined prototypes. This procedure generates a distance metric upon which visual discrimination errors are predicted or against which human errors can be correlated. PREVIP can also be used to test new visual prediction tasks and visual model improvements. Improvements to be included in future versions of PREVIP are stimulus contrast and illumination effects, masking phenomena, and peripheral visual tasks. Author