Abstract
The human visual system has been shown to be optimized to process content with the 1/f slope of the amplitude spectra of natural scenes. However, work on this topic considers processing in the context of the averaged amplitude spectrum (ignoring orientation). Recently, we have shown that visual processing of orientation in 1/f broadband content exhibits an anisotropy quite different from the traditional oblique effect obtained with simple stimuli (instead showing a “horizontal effect”). Here we examine this effect utilizing sets of natural scene imagery possessing a broad range of amplitude spectrum slopes to determine if the horizontal effect is dependent upon the slope of the amplitude spectrum of natural scenes. Sets of images containing approximately equal content (amplitude) at all orientations were gathered and filtered to be isotropic with each set containing images with amplitude spectra slopes within a narrow range. Test stimuli were generated by making an increment in the amplitude of the images at one of four 45deg. bands of orientation (0, 45, 90, or 135deg.). Ability to detect the oriented increments was measured with a single-interval Yes/No task. A control condition was also employed in which the phase spectra of the images were scrambled. Irrespective of image slope, performance for detecting oriented increments was consistently poor at horizontal and best at the obliques. The magnitude of this effect was strongest when the slope of the amplitude spectrum was most similar to that typical of natural scenes. We have proposed that this horizontal effect evolved to relatively decrease the saliency of horizontal content in natural scenes to emphasize other less-dominant content. The current results show that sensitivity for horizontal was always poor, and that the extent to which sensitivity at the other orientations was ‘emphasized’ depended on the slope of the amplitude spectrum.