Spatial-frequency model for hyperacuity

Abstract

Humans can detect vernier displacements of two abutted lines that are 30 times smaller than the bar spacings that determine their grating acuity. Since vernier acuity tasks, and hyperacuity tasks in general, reveal such drastically improved sensitivity, it has been traditionally assumed that the detection mechanisms responsible for hyperacuity are fundamentally different from those underlying ordinary spatial acuity. The need for unusual mechanisms is reinforced by the observation that hyperacuity is weakly affected by changes in suprathreshold contrast, whereas ordinary acuity is strongly influenced by contrast. Nevertheless, we argue that many hyperacuity tasks can be understood without resorting to special mechanisms. We have taken a previously developed contrast-detection model, based on spatial-frequency channels, and have applied it directly to a set of hyperacuity experiments. Hyperacuity performance is readily predicted without modification of the model. In addition, the model correctly predicts the insensitivity of hyperacuity to suprathreshold contrast as well as the measured result that moderate low-pass filtering of hyperacuity images does not significantly decrease hyperacuity performance.