Abstract
Introduction: Natural images are extraordinarily rich source of redundant information and the human visual system has evolved to exploit this redundancy. One such property is scale invariance reflected in 1/f amplitude spectrum of natural scenes. While a number of both psychophysical and physiological studies have confirmed the visual system’s tuning to 1/f scene statistics, our understanding of what aspects of variations in the amplitude spectrum are responsible for this tuning is still relatively incomplete. In this study we compare photometric (edge amplitude) versus geometric (edge density) properties of synthetic filtered images in driving visual sensitivity, perceived complexity and preference . Methods: We measured discrimination sensitivity, perceived complexity and visual preferences using 3 sets of images; greyscale (GS), thresholded (TH) and edges only (ED) all of which spanned a range of α (0.25, 0.75, 1.25, 1.75 & 2.25). Between these sets photometric structure varied while geometric properties remained stable. Thresholds for discriminating increases and decreases in α were measured by a 4AFC at each of the five reference levels (n=45), while visual preference (n=45) and visual complexity (n=40) were determined via 2AFC paired comparison procedure. The perceived complexity measurements were conducted using a larger range of image sets (GS, TH, ED, mountain cross-section & 3D terrain) and an expanded range α (0.5- 2.5). Results & Conclusion: Both discrimination sensitivity and visual preference peaked for α ∼ 1.25, confirming previous findings. Importantly, they were nearly identical for the GS, TH and ED images. Perceived complexity increased with α and similar to visual preferences scaled comparably across the different image types. The study findings provide converging evidence that visual discrimination sensitivity, visual preferences and perceived complexity are more strongly determined by the geometric structure of an image relative to their full photometric structure.