Apparent contrast and spatial frequency of local texture elements

On the synthesis of visual illusions using deep generative models

Visual illusions expand our understanding of the visual system by imposing constraints in the models in two different ways: i) visual illusions for humans should induce equivalent illusions in the model, and ii) illusions synthesized from the model should be compelling for human viewers too. These constraints are alternative strategies to find good vision models. Following the first research strategy, recent studies have shown that artificial neural network architectures also have human-like illusory percepts when stimulated with classical hand-crafted stimuli designed to fool humans. In this work we focus on the second (less explored) strategy: we propose a framework to synthesize new visual illusions using the optimization abilities of current automatic differentiation techniques. The proposed framework can be used with classical vision models as well as with more recent artificial neural network architectures. This framework, validated by psychophysical experiments, can be used to study the difference between a vision model and the actual human perception and to optimize the vision model to decrease this difference.

Nonblurred regions show priority for gaze direction over spatial blur

The human eye continuously forms images of our 3D environment using a finite and dynamically changing depth of focus. Since different objects in our environment reside at different depth planes, the resulting retinal images consist of both focused and spatially blurred objects concurrently. Here, we wanted to measure what effect such a mixed visual diet may have on the pattern of eye movements. For that, we have constructed composite stimuli, each containing an intact photograph and several progressively blurred versions of it, all arranged in a 3 × 3 square array and presented simultaneously as a single image. We have measured eye movements for 7 such composite stimuli as well as for their corresponding root mean square (RMS) contrast-equated versions to control for any potential contrast variations as a result of the blurring. We have found that when observers are presented with such arrays of blurred and nonblurred images they fixate significantly more frequently on the stimulus regions that had little or no blur at all ( p < .001). A similar pattern of fixations was found for the RMS contrast-equated versions of the stimuli indicating that the observed distributions of fixations is not simply the result of variations in image contrasts due to spatial blurring. Further analysis revealed that, during each 5 second presentation, the image regions containing little or no spatial blur were fixated first while other regions with larger amounts of blur were fixated later, if fixated at all. The results contribute to the increasing list of stimulus parameters that affect patterns of eye movements during scene perception.

Multiresolution wavelet framework models brightness induction effects

A new multiresolution wavelet model is presented here, which accounts for brightness assimilation and contrast effects in a unified framework, and includes known psychophysical and physiological attributes of the primate visual system (such as spatial frequency channels, oriented receptive fields, contrast sensitivity function, contrast non-linearities, and a unified set of parameters). Like other low-level models, such as the ODOG model [Blakeslee, B., & McCourt, M. E. (1999). A multiscale spatial filtering account of the white effect, simultaneous brightness contrast and grating induction. Vision Research, 39, 4361-4377], this formulation reproduces visual effects such as simultaneous contrast, the White effect, grating induction, the Todorović effect, Mach bands, the Chevreul effect and the Adelson-Logvinenko tile effects, but it also reproduces other previously unexplained effects such as the dungeon illusion, all using a single set of parameters.

The perceived contrast of texture patches embedded in natural images

The visibility of an isolated simple stimulus is known to depend on its contrast. However, when such a stimulus is surrounded by other geometrically-simple stimuli, its perceived contrast can change markedly. Here, we examined whether such effects contribute to our perception of contrasts when we view real world scenes. We show that the perceived contrast of a luminance texture patch is suppressed when it is surrounded by images of real world scenes. We also show that the amount of this suppression depends on the spatial statistics of the surrounding images. We manipulated the second-order statistics of the images and found minimal suppression of perceived contrast at "un-natural" image statistics and maximal suppression at the characteristic statistics of natural images. This suggests that contrast gain control mechanisms in our visual system are optimally engaged when we view real world images.

Investigating local network interactions underlying first- and second-order processing

We compared the spatial lateral interactions for first-order cues to those for second-order cues, and investigated spatial interactions between these two types of cues. We measured the apparent modulation depth of a target Gabor at fixation, in the presence and the absence of horizontally flanking Gabors. The Gabors' gratings were either added to (first-order) or multiplied with (second-order) binary 2-D noise. Apparent "contrast" or modulation depth (i.e., the perceived difference between the high and low luminance regions for the first-order stimulus, or between the high and low contrast regions for the second-order stimulus) was measured with a modulation depth-matching paradigm. For each observer, the first- and second-order Gabors were equated for apparent modulation depth without the flankers. Our results indicate that at the smallest inter-element spacing, the perceived reduction in modulation depth is significantly smaller for the second-order than for the first-order stimuli. Further, lateral interactions operate over shorter distances and the spatial frequency and orientation tuning of the suppression effect are broader for second- than first-order stimuli. Finally, first- and second-order information interact in an asymmetrical fashion; second-order flankers do not reduce the apparent modulation depth of the first-order target, whilst first-order flankers reduce the apparent modulation depth of the second-order target.

Lateral interactions in amblyopia

We studied lateral neural interactions in strabismic (n=6) and anisometropic amblyopes (n=3) by measuring reductions in the perceived contrast of a foveally viewed Gabor centred in a horizontal array of closely neighboring Gabors. Strabismic amblyopes, but not anisometropic amblyopes, failed to show the reduction in perceived contrast typical of normal vision [J. Opt. Soc. Amer. A 15 (1998) 1733] when lateral contrast information is available at the same orientation and spatial frequency. The strabismic amblyopes also severely misperceive the regularity of the array of Gabors flanking the test stimulus. A normal eye could model the anomalous contrast perception of the amblyopic eye, by adding an equivalent amount of spatial distortion to the stimulus. The relationship between the observed anomalies for local contrast gain control and positional sensitivity is discussed.

Modulation of perceived contrast by a moving surround

The apparent contrast of a center pattern depends on the contrast of its surround. To examine the suprathreshold perception of moving patterns, we measured the perceived contrast of a moving grating while the direction and speed of the surround patterns varied. Subjects matched the apparent contrast of a center patch embedded in surround patches to that of a patch with no surround pattern. Temporal frequency, Michelson contrast and movement direction of both center and surround patterns varied systematically. We found that: (1) contrast reduction is most prominent when the center and surround have the same velocity (velocity selectivity); (2) contrast enhancement occurs when the surround moves at a higher speed than the center, if the difference in temporal frequencies of center and surround exceeds 10-20, independent of the directional relationship between center and surround; (3) contrast reduction is stronger for higher surround contrasts with lower center contrasts; and (4) contrast enhancement is relatively unaffected by center and surround contrasts. We conclude that the contrast perception of moving patterns is influenced by directionally-selective mechanisms except at high temporal frequencies. Our results further suggest that there is not only the lateral inhibition often assumed to influence contrast gain control, but also an excitatory connection between motion encoding units.

Role of chromaticity, contrast, and local orientation cues in the perception of density

We compared the role of the red-green, blue-yellow, and luminance post-receptoral mechanisms in the perception of density. The task requires the comparison of densities between two stimuli composed of oriented bandpass elements, pseudo-randomly scattered across an area of constant size. The perception of density differences was measured by a temporal 2AFC procedure for all pairs of mechanisms and for four possible densities. We found that stimuli of identical physical densities are not perceived equally: there is a consistent bias in favour of blue-yellow stimuli which are perceived as significantly more dense than red-green and achromatic stimuli. We considered three factors that could have differentially affected the density perception of blue-yellow stimuli: an increase in the perceived size of the individual blue-yellow elements, a perceived contrast difference, and the presence of local orientation cues. We found that the increased perceived density of the blue-yellow stimuli occurred despite the fact that there was no increase in perceived size of the individual elements, and remained despite corrections for the two other factors. We conclude that the significant increase in perceived density for the blue-yellow mechanism is a global effect, associated with a perceived colour 'melting' of the elements in the array. Our data were fitted with the occupancy model of Allik and Tuulmets (1991, Perception & Psychophysics 49 303-314) and we found that blue-yellow stimuli have a greater 'occupancy' than red-green or achromatic stimuli.

Surround modulation in human vision unmasked by masking experiments

The responses of neurons in cat and monkey primary visual cortex are modulated by stimuli outside the classical receptive field. Here we report psychophysical evidence from masking experiments for two distinct types of surround modulation, one narrowly tuned to iso-orientation (stimuli with center and surround at the same orientation) and the other broadly tuned to cross-orientation (center and surround at perpendicular orientations). Surround modulation at iso- and cross-orientations showed distinct contrast dependencies, and high-contrast cross-oriented surrounds were able to completely eliminate masking. Surround modulation was modeled by subtracting divisive inhibition that raised the gain of spatial filters.

Measurement of the texture-coherence limit for bandpass arrays

Casual observation suggests that when the elements of a visual array are packed at a sufficiently high density they cohere to generate the percept of a texture. This 'texture-coherence limit' has been quantified by using arrays composed of Gabor functions, sixth Gaussian derivatives, or differences of Gaussians. In all cases the texture-coherence limit was a power-law function of the size of the elements as quantified by their space constants with an exponent averaging 0.7. Furthermore, the texture-coherence limit was independent of both element spatial frequency and contrast over a considerable range. A quantitative fit to the data is provided by a model in which the texture-coherence limit is determined by activation of complex cells, which pool a spatial range of subunit inputs, throughout the stimulus region. Possible extensions to two dimensions are considered.