Perceptual Organization: Motion, texture

Talk Session: Sunday, May 21, 2023, 8:15 – 9:45 am, Talk Room 1
Moderator: Michael Morgan, City, University of London

Talk 1, 8:15 am, 31.11

Extent of the “fading mirror” phenomenon as a function of image statistics of the ground texture for mirror placement

Kazushi Maruya1 (), Tomoko Ohtani2; 1NTT Communication Science Laboratories, 2Meiji University

When looking at a mirror from diagonally above it, the mirror may disappear. This disappearance can be a hazard when walking; however, it also has the potential for device applications and could be used as a form of optical camouflage. The stability of this "fading mirror" phenomenon depends on the texture pattern of the ground surface on which the mirror is placed. The disappearance of a mirror is not instantaneous but begins with the disappearance of a part of the mirror; the spatial area of disappearance increases over the span of a few seconds. In this study, to investigate the basic characteristics of this disappearance phenomenon, we measured the extent and speed of disappearance when a mirror (6.4(V) x 8 (H) arc deg) was placed on different ground patterns (size: 38.4 (V) x 21.6(H) arc deg). In Experiment 1, we measured the magnitude and speed of disappearance when the mirror was located on one of several texture patterns comprising bandpass filtering noise while the center frequency of spatial filtering was varied between 0.2 -6.3 c/deg. The results revealed that the fading was the fastest (several arc deg/sec) and strongest (near complete disappearance of objects) when the center frequency was approximately 1.6 c/deg. These results indicate that the fading of a mirror is triggered by a local (within several tens arc min as a diameter) suppression of edge processing, which then extends along the contour of the mirror. In Experiment 2, wherein we measured the fading extent and speed on various naturalistic images, certain image patterns (grass, water-surface, etc.) caused fading twice as fast as the noise patterns. These results suggest that fading rate prediction could improve based on the frequency spectrum of ground textures and that other factors in the higher-order image statistics may boost this fading.

Talk 2, 8:30 am, 31.12

Spatial Mechanisms Mediating Visual Responses to Symmetries in Textures

Yara Iskandar1 (), Christopher Lee1,3, Sebastian Bosse2, Peter J Kohler1; 1York University, Toronto, ON, Canada, 2Fraunhofer HHI, Germany, 3Image Engine Design Inc., Vancouver, BC, Canada

Symmetries are present at many scales in natural scenes. Humans and other animals are highly sensitive to visual symmetry, and symmetry has been shown to play a role in numerous domains of visual perception. Brain imaging studies have demonstrated that several regions in visual cortex exhibit robust and precise responses to symmetry. The current study explored the mechanisms underlying these responses, by measuring Steady-State Visual Evoked Potentials (SSVEPs) using high-density electroencephalography. Our stimuli were a class of regular textures, known as wallpaper groups: 17 unique combinations of symmetry types that represent the complete set of symmetries in 2D images. We focused on wallpaper groups PMM, which contains bilateral reflection symmetry, and P4, which contains four-fold rotation symmetry. Our SSVEP approach allows us to measure brain responses that are specific to the symmetries within each group. We measured these responses in two experiments, one (n=40) testing the influence of spatial frequency content and another (n=14) testing the influence of the repeating lattice structure that tiles the plane in all wallpaper groups. Exemplars for the spatial frequency experiment were generated based on log-domain band-limited random noise patches with center frequencies between 1 and 8 cycles-per-degree. For the lattice experiment, spatial frequency was kept constant at 2 cycles-per-degree and the ratio of the lattice to the overall wallpaper area varied between 1/12 and 1/2. Symmetry-specific responses were weaker overall for rotation compared to reflection, consistent with prior studies, but the manipulations had broadly similar effects for both: Responses were strongest at low spatial frequencies and weakened rapidly with increasing frequencies. The lattice manipulation had less dramatic effects, but results suggest that responses are stronger at lower ratios. Responses to reflection and rotation may thus depend on a similar mechanism that is highly dependent on spatial frequency and benefits from a repeating lattice structure.

Acknowledgements: This work was supported by the Vision Science to Applications (VISTA) program funded by the Canada First Research Excellence Fund (CFREF, 2016–2023) and by a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada awarded to PJK.

Talk 3, 8:45 am, 31.13

A replication and reanalysis of a classic texture segmentation study

Maria Kon1 (), Gregory Francis1; 1Purdue University

An image from Beck (1966) has been influential in textbook presentations of texture segmentation and has driven development of theories of texture segmentation and computational models of grouping. The image consists of left, middle, and right regions, where each region contains 36 copies of a backwards letter L, T, or slanted T, respectively, in the same relative locations. The reader is told that people indicate weaker segmentation between the L and T regions than between the T and slanted T regions, thereby indicating that texture segmentation is not based on shape similarity. More generally, the Beck (1966) study investigated the roles of orientation and similarity on texture segmentation by varying letter type and orientation. In Experiment 1 the task was to indicate which side of an image had the “more natural” break. For Experiment 2, participants viewed two of the regions and rated their similarity using a 10-point scale. Beck concluded that reported segmentation was based on common orientation and not similarity. Due to the seminal role of Experiment 1 and its small sample size (n=16), we initially planned to replicate Experiment 1 only using a larger sample size (n=101). The experiment only partially replicated (r=0.6). Two issues made the experiment difficult to replicate: the example stimuli and textual description are incompatible, and it seems that spacing between elements was not properly controlled in the original study. Additionally, a reanalysis finds that the data in Beck (1966) actually does suggest a role of similarity in texture segmentation: more similar regions tend to be the chosen segmentation (r=-0.7). Since this conclusion is counterintuitive, we also successfully replicated Experiment 2 (n=23, r=0.9). The reported segmentations in our Experiment 1 were partly explained (r=0.4) by the similarity ratings in our Experiment 2: more similar regions were less likely to be the chosen segmentation.

Acknowledgements: GF was supported by the European Union’s Horizon 2020 Framework Programme for Research and Innovation under the Specific Grant Agreement No. 945539 (Human Brain Project SGA3) and by a Visiting Scientist Grant from the Swiss National Science Foundation.

Talk 4, 9:00 am, 31.14

Distinct rules for binding in position-based and velocity-based motion systems

Ilker Duymaz1 (), Nihan Alp1; 1Sabanci University, Istanbul, Turkey

Motion perception depends on at least two distinct systems: A velocity-based motion system (VMS) facilitated by early direction-selective cells, and a position-based system (PMS) by tracking objects/features across space and time. However, it is still unclear how these two systems interact to produce coherent motion perception when they work in parallel. Here we investigate how both systems contribute to the perceptual organization of motion cues. To this end, we apply two manipulations to a bistable stimulus, in which eight moving dots could either be perceived as rotating in tandem with their most proximate pair –local motion– or moving around the fixation cross as the corners of two illusory squares –global motion–. The first manipulation systematically disrupts VMS by stroboscopically changing the durations of inter-frame intervals (IFIs: between 0 and 116.69 ms in steps of 16.67). This deactivates the involvement of direction-selective cells, which are known to have short temporal integration windows (<100 ms), undermining VMS. The second manipulation enhances different grouping configurations (locally-grouped, globally-grouped, and no-group) by applying contrast polarity. This allows comparing the effectiveness of the static grouping cues (i.e., proximity and contrast) when only the PMS is active as opposed to when both systems are active. We find that the pattern of bistability changes substantially at IFIs of 33 ms. For IFIs ≥33 ms, perceptual bistability is heavily biased by the static grouping cues. For IFIs <33 ms, perceptual bistability favors the global percept over the local even when the static cues enhance the local grouping, indicating that the contribution of VMS counteracts static grouping cues. These results reveal that distinct rules in VMS and PMS govern the perceptual organization of moving stimuli and that conflicting perceptual configurations between the two systems might explain bistable motion perception.

Acknowledgements: This work is funded by TÜBİTAK 3501 Career Development Grant (220K038).

Talk 5, 9:15 am, 31.15

Neural correlates of perceptual motion integration and segmentation of locally paired and unpaired random-dot stimuli.

Bikalpa Ghimire1 (), Xin Huang1; 1University of Wisconsin- Madison

Primate visual system can integrate local motion vectors into a global motion pattern and segment multiple visual stimuli based on motion cues. Previous psychophysical studies showed that when random dots of overlapping stimuli moving in two directions were locally paired, human subjects perceived an integrated vector-averaged (VA) direction of the stimuli; whereas when the dots were unpaired, subjects perceived motion transparency and two component directions. The mechanism underlying this drastic perceptual change is unknown. We recorded from neurons in the middle-temporal (MT) cortex of fixating monkeys. Visual stimuli were random dots moving at 5⁰/s in two directions separated by 90⁰. We varied the VA direction of the stimuli to characterize the direction tuning. In the “paired-dot” condition, two dots moving in different directions were locally paired within a path of 0.4⁰ (lifetime of 80 ms). In the “unpaired-dot” condition, dots moving in two directions were unpaired and had the same lifetime. We found that in response to the unpaired-dot stimuli, neurons showed bimodal tuning which represented two component directions, similar to our previous finding using stimuli that had a much longer lifetime and elicited strong motion transparency. Remarkably, MT neurons showed a unimodal tuning to the paired-dot stimuli and the peak was reached when the VA direction of the stimuli was near a neuron’s preferred direction. The shape of the unimodal tuning closely matched the tuning to single motion directions, suggesting MT responses to the paired-dot stimuli correlate with perceptual integration. Moreover, when the paired-dot stimuli were placed closer to fovea, they appeared to contain two directions and the MT tuning changed from unimodal to bimodal, again correlated with the perceptual change. Our findings reveal neural correlates of an intriguing perceptual phenomenon and have implications for the roles of local computation in early visual areas on motion integration and segmentation.

Acknowledgements: This study is supported by NIH grant R01EY022443 from the National Eye Institute.

Talk 6, 9:30 am, 31.16

Object solidity disambiguates ambiguous motion

Dawei Bai1 (), Brent Strickland1,2; 1École Normale Supérieure, PSL Research University, Institut Jean Nicod (ENS, EHESS, CNRS), Paris, France, 2Africa Business School; School of Collective Intelligence - UM6P - Rabat, Morocco

The human visual system is endowed with expectations about physical regularities it routinely perceives. One basic regularity of our world is that of “object solidity”: Solid objects cannot pass through each other. Surprisingly, not only is there no convincing evidence that solidity is embedded in automatic visual processes, the visual system has been shown to tolerate obvious violations of this physical constraint (Ames, 1951; Leslie, 1988; Wilson & Robinson, 1986). Here we provide novel evidence that the adult human visual system does in fact use solidity to guide perception – in particular interpretations of ambiguous motion. When viewing displays compatible with multiple interpretations, participants consistently favored interpretations which did not violate object solidity over those that did (for a demo, see Specifically, two ambiguously rotating rings that partially overlapped were predominantly perceived as moving in 180° rotation (which respects solidity), instead of 360° rotation (where the rings traverse each other). This preference diminished or disappeared if the rings were separated or holed, and could not be overridden by a contradictory disambiguating cue – 360° rotating shadows. Further demonstrating the robustness of this effect, we discovered that a linearly moving circle was more likely perceived as moving circularly in depth when a vertical bar was introduced in its path. This effect persisted even if the circle varied in size during its motion (as would a circling ball). Together, our results demonstrate that the visual system integrates solidity in its computations of ambiguous motions, preferring the interpretations respecting solidity.

Acknowledgements: This research received support from PSL University (IPFBW 2016-151), Agence Nationale de la Recherche (grants ANR-10-IDEX-0001-02 PSL and ANR-17-EURE-0017 FrontCog) and the European Union (FP/2007-2013; 324115, FRONTSEM; 778077, Orisem).