Effect of image orientation contents on detection efficiency.

Abstract

Contrast detection performance is known to be better for single component sinusoidal gratings than for sums of gratings at different orientations. A recent study Rovamo et al. (1994) (Investigative Ophthalmology and Visual Science, 35, 2611-2619) showed that spatial integration is less effective for multiple orientation component than for single component gratings. This suggests an explanation that the size of a spatial integration window depends on the orientation contents of the stimulus. To test this hypothesis we designed a computational detection model and tested it against new experimental data. The model generates a cross-correlation template, the extent of which is limited both in the spatial and spatial frequency domain. The template is a copy of the band-pass filtered signal weighted by a spatial window function. The spatial window function, which limits spatial integration, decreases with increasing orientation range of the stimulus. The experimental stimuli were composed of side-by-side located square shaped, one cycle, grating patches. The range of either grating orientations or phases within the patches as well as the number of patches in a stimulus were varied. We also measured detection efficiency for Bessel Jo images as a function of area. Human spatial integration became considerably weaker with increasing orientation range. The increasing phase range also reduced detection efficiency to some extent. Supporting the idea of the varying size of the spatial integration window, the computational model explained the orientation, phase, and Bessel Jo data well.