3D perception II
The Herringbone Depth Effect: The Tower of Toulouse
The effect reported here was first observed after several hundred vision scientists attending the 2011 ECVP reception left the main hall of the Capitolium in Toulouse, France. The floor of the hall is tiled with a wooden herringbone pattern. Without the visual barrier produced by the people filling the hall, the floor immediately took on a “corrugated” appearance. Moreover, a quick photograph taken with a cell phone revealed the “Tower of Toulouse” in which a half dozen of the long herringbone lanes on the floor appeared to “stand up” into a well-defined tower-like structure. A comparable visual effect is elicited by a bent Mach card, which can appear to be either horizontal or vertical when lying on a horizontal surface. The Mach card, however, is an actual three-dimensional structure. In the new effect reported here, the two-dimensional floor gives rise to a very robust three-dimensional “tower”. The herringbone pattern has been used in floor tilings for over two millennia. Pliny the Elder in his “Natruralis Historia” - written before 79 AD - described “tile floors with a herringbone pattern”. In the intervening two millennia, there appears to have been no report of this effect in the vision science literature.
Depicting the visual field in art and science
This paper will discuss my attempts to depict the visual field in painting and drawing. Depicting visual the visual field means trying to capture natural scenes as they are actually perceived rather than as they might appear in, say, a photograph or computer generated rendering. As I will show, there are a number of fundamental features of visual perception that conventional imaging technology does not record. The most important of these are the differentiation between central and peripheral vision and the relative indeterminacy of objects in the periphery, deformations of objects in space relative to viewing position, and the presence of the viewer's own body in the field of view. Once these features are accommodated into the depiction, I argue, we arrive at image that is much closer to actual visual perception than images that conform to linear perspective and omit the appearance of the viewer's own body in the periphery of the visual field. The paper will consider the implications of this approach for the scientific study of perception, how it links to some recent neuroscientific research, and how artists and scientists might benefit from further developing the methods outlined here.
Distinct correlates of reversing illusory rotation or depth for the structure-from-motion: an MEG study
Alexander Pastukhov, Mandy Bartsch, Solveiga Stonkute, Jens-Max Hopf and Jochen Braun
A cloud of dots in planar motion can induce the compelling illusion of a rotation in depth (“structure-from-motion”). Surprisingly, reversing the planar motion does not necessarily reverse the global illusory rotation. Alternatively, the dots may locally reverse illusory depth, thus allowing the global illusory rotation to continue. This ambiguous outcome is obtained only if the global illusory shape remains unchanged [Pastukhov et al, 2012, Journal of Vision 12(1):17]. These psychophysical observations imply that global illusory shape is represented independently of a local illusory depth. Here we use EEG/MEG recording to compare the neural correlates of these alternative perceptual interpretations (reversal of global illusory rotation, GIR, or of local illusory depth, LID) of a physically identical event (reversal of planar motion). Reversals of GIR are associated with phasic activity in a medial-temporal area (presumptive hMT, ~180 ms) and, later, in an immediately adjacent superior region (~270 ms). In contrast, reversals of LID are associated with phasic activity near the intraparietal sulcus (presumptive LIP) and in ventral extrastriate areas (~215 ms). We interpret the latter activity pattern as reflecting renewed “binding” between neural representations of local depth and of global rotation.
Recovering 3-D shape: Roles of absolute and relative disparity, retinal size, and vergence as revealed by reverse-perspective stimuli
Thomas V. Papathomas, Joshua Dobias, Daniel Moritz and Geetika Baghel
Purpose: Investigate roles of stimulus size, binocular disparity and vergence under key conditions. Reverspective stimuli were selected because data-driven signals (disparity, motion parallax, etc.) help recover veridical 3D shape; they compete against schema-driven influences (experience with perspective, foreshortening, pictorial cues) favoring the illusory depth inversion. The ensuing dynamic depth reversals help study depth perception. Methods: Three scaled-size versions of a reverspective were used. The viewing distance in three conditions was varied while keeping one parameter fixed across the three stimuli: (1) fixed retinal size, (2) fixed disparity, (3) fixed vergence angle. The predominance of the veridical percept was recorded. Results: Unexpectedly, the illusion strength was the same when the vergence was fixed, despite significantly different disparities and retinal sizes; conversely, illusion strength changed significantly in fixed disparity and fixed retinal size conditions. The illusion was stronger for greater distances or, equivalently, for smaller vergence angles. “Relative disparity” (disparity normalized through division by stimulus size) was a good predictor of the data trends. Conclusions: Two possible explanations for the results: (1) Vergence may play a large role in resolving the perceptual conflict between disparity and perspective cues. (2) Disparities may be normalized by stimulus size to recover 3-D shape.
Time in Perspective
Andrei Gorea and Janice Hau
Perceived size of a constant retinal size object increases as it is displaced toward the vanishing point of a 2D-scene rendered in linear perspective (Ponzo-illusion). The perceived duration (PeDu) of a moving object increases with the trajectory it covers (Kappa-effect; Abbe, 1936). The two phenomena lead to predict that the PeDu taken by moving objects placed in the background of a linear perspective scene will be longer than the PeDu taken by the same objects moving at the same speeds and covering the same angular trajectories in the foreground. Using bicolored 3D-rendered balls rolling in a fronto-parallel plane of a linear perspective 2D-scene (checkerboard), we show that psychological time flows up to 50% faster as the fronto-parallel plane recedes from the observer. Such PeDu dilation was confirmed for 3 physical durations (600, 900, 1200 ms) and 3 trajectory lengths (5.5, 11, 22 deg). Control experiments show that the phenomenon is contributed to in different proportions by the Ponzo-Kappa combination, the relative balls’ sizes, the relative sizes of the background checkers, and the perspective cues (vanishing point) proper. For any of these reasons or for all of them, psychological time expands with the (apparent) distance from the observer.
From Orientation Flows to Surface Inferences
Steven Zucker, Benjamin Kunsberg and Roland Fleming
One of the most important functions of vision is to estimate the 3D shape of objects in our environment. Many different cues (e.g. disparities, shading, texture) provide information about shape, but how the visual system estimates shape is poorly understood. It is well understood, however, that the (early) visual system is organized around orientation. Here we present evidence that crucial information is extracted from the way local image orientation signals vary continuously across the surface of an object ('orientation flows'), and that these flows provide the foundation for surface inferences. To start, striking regularities in the flows emerge when computer renderings of shaded and textured objects are represented in a (superficial-layer) V1 fashion. These orientation flows change when illumination and texture patterns change, leading to a number of psychophysical predictions. A model of shape inference from shading flows reveals how surface and light source properties emerge from the flows, and the geometry of the model could be learned by the visual system. Together these findings suggest that the visual estimation of shape from shading, highlights and texture may have more in common than previously thought, and that orientation fields could act as a 'common currency' for the visual estimation of shape.
Stereomotion scotomas in healthy subjects: stability and cue-dependence
Martijn Barendregt, Serge Dumoulin and Bas Rokers
The visual field of many people contains regions with decreased or no sensitivity to stereomotion, called stereomotion scotomas (Hong & Regan, 1989). Recently, Rokers et al (2008) revealed two independent binocular cues underlying stereomotion: interocular velocities and changing disparity over time. We hypothesized that stereomotion scotomas are cue-specific. To test our hypothesis, we measured sensitivity to stereomotion across the visual field in 7 subjects. Because the two stereomotion cues have different speed tuning curves (Czuba et al, 2010), we used a stimulus (Gabor, sigma: 0.5deg, lambda: 4cpd) with different monocular speeds to isolate the sensitivity to each cue (changing disparity: 0.25 deg/s - slow, and interocular velocity 2 deg/s - fast). There was a significant difference between the performance in the slower and faster speed condition (mean 66.1% and 79.8% correct, respectively). We found evidence for stereomotion scotomas in both the slow condition (2/7 subjects) and fast condition (3/7), which did not warrant conclusions about the underlying cue. We did reveal that the stereomotion field sensitivities were stable over time within each speed condition (mean correlations 48.1% [slow] and 55.9% [fast]), and furthermore that the stereomotion scotomas could not be explained by a drop off of performance with eccentricity.
Missperception helps the action – anisotropy of perceived distance and effort
Distances towards zenith are perceived longer than physically same distances towards horizon. We argued that this kind of perceived distance anisotropy is in a function of action. Namely, if one tends to reach something upwards, opposite to gravity direction, more effort is needed. If visual system overestimates distance, reaching for something further away would demand more effort, and action opposite to gravity direction would be easily done. We tested this hypothesis in two experiments. In a first experiment 15 participants had a task to equalize the distances of two stimuli by hand. Stimuli were on horizontal and vertical direction, on 0.2m, 0.4m and 0.6m distances. Results have shown that participants matched shorter vertical with longer horizontal distances, meaning that they perceived vertical distances as longer. These results show that same anisotropy exists in proprioceptive space as in visual space. In a second experiment 14 participants matched efforts of 2kg, 4kg and 6kg by stretching dynamometer, on horizontal and vertical direction. Results have shown that participants tend to match larger horizontal efforts with smaller vertical, meaning that they perceived vertical efforts as more intensive. Results from both experiments are in line with hypothesis that anisotropy is in a function of action. This research was supported by Ministry of education and science, Republic of Serbia, project number 179033.