Sampling irregularity perturbs visual reconstruction.

Abstract

We explored the human observer's ability to detect and discriminate sine-wave and square-wave gratings that were sampled at intervals varying from 4.7 to 9.4 arcmin. To study the effect of sampling irregularity on visual performance, we varied the position of each line sample on the basis of a Gaussian probability distribution, the standard deviation of which varied from 0 (regular sampling) to 4.7 arcmin (highly irregular sampling). The results indicate that irregular sampling has no systematic effect on the observer's ability merely to detect the presence of a sine- or square-wave grating. In contrast, sampling irregularity strongly impairs the subject's ability to discriminate between these waveforms. A model based on the convolution of difference-of-Gaussians-type weighting profiles predicted that sampling irregularity should have little to no effect on the output of a channel tuned to the third harmonic of the square-wave grating. The findings thus suggest the existence of a sampling scheme in the visual system. This scheme is based on local feature-selective mechanisms, probably edge detectors, that are highly sensitive to the relative position of the sample points in the space domain.