What's constant in contrast constancy? The effects of scaling on the perceived contrast of bandpass patterns

Abstract

"Contrast constancy" refers to the ability to perceive objects as maintaining a constant contrast independent of size or distance. When tested with high contrast sinusoidal gratings, contrast constancy has been shown to hold for a wide range of spatial frequencies, suggesting that sensitivity is constant across the spectrum at suprathreshold. In this study, we show that contrast constancy also holds for relatively broadband patterns. We describe how the frequency spectra of such functions change as the patterns scale in size. In particular, we emphasize how these changes in the spectra depend on whether the functions are localized (coherent phase) or spatially distributed (incoherent phase). In Fourier terms, the scaling properties depend on the phase spectra of the patterns. Contrast constancy is shown to hold for both localized Gabor patches (coherent phase spectra) and bandpass noise patterns (incoherent phase spectra). Constancy holds over a wide range of suprathreshold contrasts; in fact, matching is quite accurate as soon as the pattern is suprathreshold. These results are explained with a model in which mechanism bandwidths increase with frequency (constant in octaves) and peak spectral sensitivity is equal across frequency out to around 16 c/deg. In the case of the Gabor stimuli, perceived contrast is assumed to be mediated by a mechanism centered on the patch. For the bandpass noise, contrast is determined by the average response of units distributed across the stimulus. This model can account for the matching data without assuming that the contrast-response gain of the underlying channels changes with spatial frequency. Neither does the model assume "response pooling". In addition to explaining the experimental results, the model also predicts that perceived contrast will be approximately constant across scale for scenes whose spectra fall as 1/f, as is typical of natural scenes.

A multiscale model of adaptation and spatial vision for realistic image display

“…Finally, the rod system is processed through a distinct set of transducers to represent its unique spatial characteristics. At high contrast levels (> 5%) the transducer functions converge to a common square-root form to properly represent perceived contrast constancy [1] and introduce a compressive nonlinearity typically found in masking experiments and color appearance models. The contrast transducers used in our model are illustrated in Figure 3.…”

Reminder and 2AFC tasks provide similar estimates of the difference limen: A reanalysis of data from Lapid, Ulrich, and Rammsayer (2008) and a discussion of Ulrich and Vorberg (2009)

“…Research on visual illusions (see, e.g., Armstrong & Marks, 1997;Chacón, 2004;Foley & McCourt, 1985;Jordan & Schiano, 1986) has often used discrimination tasks to provide objective data (i.e., beyond the initial consensus of introspections), indicating that the illusion actually occurs when the PSE is located at the point at which standard and comparison stimuli have the same magnitude along the dimension in which they are compared. Similar evidence that the PSE and the POE do not generally coincide comes also from other common studies involving nonillusory standard and comparison stimuli that differ along noncompared dimensions, like windowed sinusoidal gratings of different orientation, mean luminance, direction of motion, spatial frequency, size, or spatial phase whose contrast is to be compared (see, e.g., Brady & Field, 1995; García-Pérez know whether the functions S and T are identical even when test and standard stimuli are nominally identical in all respects (see our earlier discussion of temporary sensory impairment). In these conditions, the only valid conclusion from Ulrich and Vorberg's analysis is that 2AFC (x) .5 at some point x, which may or may not be the point x s; then, the PSE may or may not lie at the POE.…”

What determines the maximum displacement limit for spatially broadband kinematograms?

“…This means that its energy spectrum needs to follow a 1͞f 2 slope. 25 Examples of a RDP and the same pattern scaled in the Fourier domain by a 1͞f transfer function to yield what will be termed here a fractal-noise pattern (FNP) (with a 1͞f 2 energy spectrum) are shown in Fig. 1.…”