On the parameters of human visual performance: an investigation of the benefits of antialiasing

Abstract

A two-part experiment was conducted to investigate the effects of aliasing artifacts and screen resolution on a simple visual recognition task. The results indicate that in many cases far less realism may be necessary in synthetic computer-generated imagery than is often assumed in the literature. The first part of the experiment comprised a subjective rating of image quality, the second part measured task effectiveness of image quality. In the second part subjects were asked to discriminate between images of two types of objects built from cubes, similar to objects used in experiments involving mental rotation. At higher resolutions the elimination of aliasing artifacts did not significantly improve subjects' performance. At intermediate and low resolutions, comparable to what might be used for iconic menus, the reduction in aliasing artifacts resulted in improved performance. The subjective ratings indicate that for both high and low resolution the elimination of aliasing artifacts does not improve “quality,” whereas images rendered at intermediate resolutions are significantly degraded by aliasing artifacts to the extent that antialiasing improves the subjective rating. An interpretation of these results is given in the context of an ongoing research program aimed at identifying the parameters of real-time human performance for graphics workstations.