Illusory Contours Research Articles

Emergence of illusory contours in robust deep neural networks by accumulation of implicit priors

Emergence of illusory contours in robust deep neural networks by accumulation of implicit priors

Visual search of illusory contours: The role of illusory contour clarity.

Illusory contours demonstrate an important function of the visual system-object inference from incomplete boundaries, which can arise from factors such as low luminance, camouflage, or occlusion. Illusory contours can be perceived with varying degrees of clarity depending on the features of their inducers. The present study aimed to evaluate whether illusory contour clarity influences visual search efficiency. Experiment 1 compared visual search performance for Kanizsa illusory stimuli and nonillusory inducer stimuli when manipulating inducer size as a clarity factor. Experiment 2 examined the effects of illusory contour clarity on visual search by manipulating the number of rings with missing arcs (i.e., line ends) comprising the inducers, for both illusory and nonillusory stimuli. To investigate whether surface alterations had an impact on visual search in Experiment 1, Experiment 3 examined search performance for Kanizsa-like stimuli formed from "smoothed" inducers compared with standard Kanizsa figures. The results of Experiments 1 and 2 indicated that while Kanizsa produced inefficient search, this was not contingent on the clarity of the illusory contours. Experiment 3 suggested that surface alterations of Kanizsa figures did impact visual search performance. Together, the results indicated that illusory contour clarity did not have much bearing on search performance. In certain conditions, Kanizsa figures even facilitated search compared with nonillusory stimuli, suggesting that rather than contour inference, surface features might have greater relevance in guiding visual attention.

It's all in the timing: delayed feedback in autism may weaken predictive mechanisms during contour integration.

Humans rely on predictive and integrative mechanisms during visual processing to efficiently resolve incomplete or ambiguous sensory signals. Although initial low-level sensory data are conveyed by feedforward connections, feedback connections are believed to shape sensory processing through automatic conveyance of statistical probabilities based on prior exposure to stimulus configurations. Individuals with autism spectrum disorder (ASD) show biases in stimulus processing toward parts rather than wholes, suggesting their sensory processing may be less shaped by statistical predictions acquired through prior exposure to global stimulus properties. Investigations of illusory contour (IC) processing in neurotypical (NT) adults have established a well-tested marker of contour integration characterized by a robust modulation of the visually evoked potential (VEP)-the IC-effect-that occurs over lateral occipital scalp during the timeframe of the visual N1 component. Converging evidence strongly supports the notion that this IC-effect indexes a signal with significant feedback contributions. Using high-density VEPs, we compared the IC-effect in 6- to 17-yr-old children with ASD (n = 32) or NT development (n = 53). Both groups of children generated an IC-effect that was equivalent in amplitude. However, the IC-effect notably onset 21 ms later in ASD, even though initial VEP afference was identical across groups. This suggests that feedforward information predominated during perceptual processing for 15% longer in ASD compared with NT children. This delay in the feedback-dependent IC-effect, in the context of known developmental differences between feedforward and feedback fibers, suggests a potential pathophysiological mechanism of visual processing in ASD, whereby ongoing stimulus processing is less shaped by visual feedback.NEW & NOTEWORTHY Children with autism often present with an atypical visual perceptual style that emphasizes parts or details over the whole. Using electroencephalography (EEG), this study identifies delays in the visual feedback from higher-order sensory brain areas to primary sensory regions. Because this type of visual feedback is thought to carry information about prior sensory experiences, individuals with autism may have difficulty efficiently using prior experience or putting together parts into a whole to help make sense of incoming new visual information. This provides empirical neural evidence to support theories of disrupted sensory perception mechanisms in autism.

Perception of illusory contours in children and adults: An eye-tracking study.

The eye-tracking study investigated the perception of subjective Kanizsa and Ehrenstein figures in adults and in children aged 3-4, 5-6, 7-8, and 9-11 years of age. More specifically, the distribution of looking at the inner stimulus part versus the inducing elements was measured for illusory figures, figures with real contours, and control displays. It was hypothesized that longer looking at the inner area of the illusory figures indicates global contour interpolation, whereas longer looking at the inducing elements indicates a local processing mode. According to the results, participants of all ages looked longer at the illusory Kanizsa and Ehrenstein contours than at the figures' inducing elements. However, performance was lowest in the children aged 3-4 years and increased during the preschool period. Moreover, the illusory contour displays elicited comparable visual responses as did the real contour displays. The use of the control displays that contained no contour information ensured that the participants' looking behavior was not driven by a spontaneous tendency to attend to the inner stimulus parts. The study confirms the view that sensitivity to illusory contours emerges very early in life.

Visual saltation illusion induced by flashes of subjective contours.

The visual saltation illusion of a Kanizsa-type subjective triangle was demonstrated. After a subjective/real triangle was flashed twice at the same position, another subjective/real triangle was flashed at a displaced position. In a typical case, the second flash was perceived to occur midway between the first and third flash positions. This study showed that the rated illusion strength for the Kanizsa and real triangles largely depended on stimulus onset asynchrony and retinal eccentricity and that the illusion rating was the same between the Kanizsa and real gray triangles when they were presented on black disks (or inducers). When the real triangle was presented in isolation, the illusion rating was lower. Presenting flashes on disks appears to enhance the saltation illusion for both the Kanizsa and real triangles possibly due to a stronger crowding effect or shape changes of inducers enhancing the perception of object appearance and disappearance. Various types of saltation illusions with a Kanizsa triangle are demonstrated in a video.

Severely Attenuated Visual Feedback Processing in Children on the Autism Spectrum.

Individuals on the autism spectrum often exhibit atypicality in their sensory perception, but the neural underpinnings of these perceptual differences remain incompletely understood. One proposed mechanism is an imbalance in higher-order feedback re-entrant inputs to early sensory cortices during sensory perception, leading to increased propensity to focus on local object features over global context. We explored this theory by measuring visual evoked potentials during contour integration as considerable work has revealed that these processes are largely driven by feedback inputs from higher-order ventral visual stream regions. We tested the hypothesis that autistic individuals would have attenuated evoked responses to illusory contours compared with neurotypical controls. Electrophysiology was acquired while 29 autistic and 31 neurotypical children (7-17 years old, inclusive of both males and females) passively viewed a random series of Kanizsa figure stimuli, each consisting of four inducers that were aligned either at random rotational angles or such that contour integration would form an illusory square. Autistic children demonstrated attenuated automatic contour integration over lateral occipital regions relative to neurotypical controls. The data are discussed in terms of the role of predictive feedback processes on perception of global stimulus features and the notion that weakened "priors" may play a role in the visual processing anomalies seen in autism.SIGNIFICANCE STATEMENT Children on the autism spectrum differ from typically developing children in many aspects of their processing of sensory stimuli. One proposed mechanism for these differences is an imbalance in higher-order feedback to primary sensory regions, leading to an increased focus on local object features rather than global context. However, systematic investigation of these feedback mechanisms remains limited. Using EEG and a visual illusion paradigm that is highly dependent on intact feedback processing, we demonstrated significant disruptions to visual feedback processing in children with autism. This provides much needed experimental evidence that advances our understanding of the contribution of feedback processing to visual perception in autism spectrum disorder.

Challenging deep learning models with image distortion based on the abutting grating illusion

Even state-of-the-art deep learning models lack fundamental abilities compared with humans. While many image distortions have been proposed to compare deep learning with humans, they depend on mathematical transformations instead of human cognitive functions. Here, we propose an image distortion based on the abutting grating illusion, which is a phenomenon discovered in humans and animals. The distortion generates illusory contour perception using line gratings abutting each other. We applied the method to MNIST, high-resolution MNIST, and "16-class-ImageNet" silhouettes. Many models, including models trained from scratch and 109 models pretrained with ImageNet or various data augmentation techniques, were tested. Our results show that abutting grating distortion is challenging even for state-of-the-art deep learning models. We discovered that DeepAugment models outperformed other pretrained models. Visualization of early layers indicates that better-performing models exhibit the endstopping property, which is consistent with neuroscience discoveries. Twenty-four human subjects classified distorted samples to validate the distortion.

Visual search of illusory contours: The role of illusory contour clarity.

Kanizsa-type illusory contours demonstrate an important function of the visual system-object inference from incomplete boundaries, which can be due to low luminance environments, camouflage, or occlusion. At a perceptual level, Kanizsa figures have been shown to have various degrees of clarity, depending on the features of the inducers. The aim of the present study is to evaluate whether contour clarity influences search efficiency of Kanizsa-type illusory contours. Experiment 1 will examine search for a Kanizsa-type illusory target among Kanizsa-type illusory distractors, by manipulating contour clarity using inducer size in three conditions, compared with search for a nonillusory perceptually grouped target among nonillusory perceptually grouped distractors with manipulated inducer size. Experiment 2 will address the effects of contour clarity on visual search by manipulating the number of arcs (i.e., line ends) comprising the inducers, in a visual search task of Kanizsa-type stimuli, compared with visual search for nonillusory grouped targets and distractors when the number of arcs are manipulated. To examine whether surface alterations had an impact on search in Experiment 1due to changes in inducer size, Experiment 3 will examine search for Kanizsa stimuli formed from "smoothed" inducers, in comparison tosearch for Kanizsa stimuli used in Experiment 1. Together, these experiments will demonstrate whether contour clarity impacts visual search of illusory contours.

EEG Signal-Assisted Algebraic Topological Feature-Enhanced Deep Neural Networks for Gestalt Illusory Contour Perception

EEG Signal-Assisted Algebraic Topological Feature-Enhanced Deep Neural Networks for Gestalt Illusory Contour Perception

Are dogs indeed susceptible to Kanizsa's triangle illusion?

Survival often depends on the ability of the visual system to process information accurately; thus, research demonstrating that a brain is susceptible to optical illusions is of considerable interest, particularly when the experiments involve phylogenetic comparisons. Are Lõoke et al.'s (Anim. Cogn, 25:43-51, 2022) data strong enough to allow the inclusion of dogs on the list of nonhumans that can perceive illusory Kanizsa figures?

The strength of feedback processing is associated with resistance to visual backward masking during Illusory Contour processing in adult humans

Re-entrant feedback processing is a key mechanism of visual object-recognition, especially under compromised viewing conditions where only sparse information is available and object features must be interpolated. Illusory Contour stimuli are commonly used in conjunction with Visual Evoked Potentials (VEP) to study these filling-in processes, with characteristic modulation of the VEP in the ∼100-150 ms timeframe associated with this re-entrant processing. Substantial inter-individual variability in timing and amplitude of feedback-related VEP modulation is observed, raising the question whether this variability might underlie inter-individual differences in the ability to form strong perceptual gestalts. Backward masking paradig ms have been used to study inter-individual variance in the ability to form robust object perceptions before processing of the mask interferes with object-recognition. Some individuals recognize objects when the time between target object and mask is extremely short, whereas others struggle to do so even at longer target-to-mask intervals. We asked whether timing and amplitude of feedback-related VEP modulations were associated with individual differences in resistance to backward masking. Participants (N=40) showed substantial performance variability in detecting Illusory Contours at intermediate target-to-mask intervals (67 ms and 117 ms), allowing us to use kmeans clustering to divide the population into four performance groups (poor, low-average, high-average, superior). There was a clear relationship between the amplitude (but not the timing) of feedback-related VEP modulation and Illusory Contour detection during backward masking. We conclude that individual differences in the strength of feedback processing in neurotypical humans lead to differences in the ability to quickly establish perceptual awareness of incomplete visual objects.

Bio-Inspired Models of Convolution Neurons in the Problem of Illusory Contour Recognition

This paper presents the result of designing the architecture of a neural network on bio-inspired neurons, whose task is to work out the mechanism for recognizing an illusory contour using the example of “Kanizsa’s figures”. The neural network made it possible to achieve invariance to the number of corners of the figure and does not lose recognition quality when changing the size of the illusory contour. The main application of the approach can be found in the problem of separating “figure-background” in images.

The complexity of simple counting: ERP findings reveal early perceptual and late numerical processes in different arrangements

The counting process can only be fully understood when taking into account the visual characteristics of the sets counted. Comparing behavioral data as well as event-related brain potentials (ERPs) evoked by different task-irrelevant arrangements of dots during an exact enumeration task, we aimed to investigate the effect of illusory contour detection on the counting process while other grouping cues like proximity were controlled and dot sparsity did not provide a cue to the numerosity of sets. Adult participants (N = 37) enumerated dots (8–12) in irregular and two different types of regular arrangements which differed in the shape of their illusory dot lattices. Enumeration speed was affected by both arrangement and magnitude. The type of arrangement influenced an early ERP negativity peaking at about 270 ms after stimulus onset, whereas numerosity only affected later ERP components (> 300 ms). We also observed that without perceptual cues, magnitude was constructed at a later stage of cognitive processing. We suggest that chunking is a prerequisite for more fluent counting which influences automatic processing (< 300 ms) during enumeration. We conclude that the procedure of exact enumeration depends on the interaction of several perceptual and numerical processes that are influenced by magnitude and arrangement.

Cortical circuits for top-down control of perceptual grouping

A fundamental characteristic of human visual perception is the ability to group together disparate elements in a scene and treat them as a single unit. The mechanisms by which humans create such groupings remain unknown, but grouping seems to play an important role in a wide variety of visual phenomena, and a good understanding of these mechanisms might provide guidance for how to improve machine vision algorithms. Here, we build on a proposal that some groupings are the result of connections in cortical area V2 that join disparate elements, thereby allowing them to be selected and segmented together. In previous instantiations of this proposal, connection formation was based on the anatomy (e.g., extent) of receptive fields, which made connection formation obligatory when the stimulus conditions stimulate the corresponding receptive fields. We now propose dynamic circuits that provide greater flexibility in the formation of connections and that allow for top-down control of perceptual grouping. With computer simulations we explain how the circuits work and show how they can account for a wide variety of Gestalt principles of perceptual grouping, texture segmentation tasks, amodal illusory contours, and ratings of perceived groupings. We propose that human observers use such top-down control to implement task-dependent connection strategies that encourage particular groupings of stimulus elements in order to promote performance on various visual tasks.

Nonsymbolic numerosity in sets with illusory-contours exploits a context-sensitive, but contrast-insensitive, visual boundary formation process

The visual mechanisms underlying approximate numerical representation are still intensely debated because numerosity information is often confounded with continuous sensory cues (e.g., texture density, area, convex hull). However, numerosity is underestimated when a few items are connected by illusory contours (ICs) lines without changing other physical cues, suggesting in turn that numerosity processing may rely on discrete visual input. Yet, in these previous works, ICs were generated by black-on-gray inducers producing an illusory brightness enhancement, which could represent a further continuous sensory confound. To rule out this possibility, we tested participants in a numerical discrimination task in which we manipulated the alignment of 0, 2, or 4 pairs of open/closed inducers and their contrast polarity. In Experiment 1, aligned open inducers had only one polarity (all black or all white) generating ICs lines brighter or darker than the gray background. In Experiment 2, open inducers had always opposite contrast polarity (one black and one white inducer) generating ICs without strong brightness enhancement. In Experiment 3, reverse-contrast inducers were aligned but closed with a line preventing ICs completion. Results showed that underestimation triggered by ICs lines was independent of inducer contrast polarity in both Experiment 1 and Experiment 2, whereas no underestimation was found in Experiment 3. Taken together, these results suggest that mere brightness enhancement is not the primary cause of the numerosity underestimation induced by ICs lines. Rather, a boundary formation mechanism insensitive to contrast polarity may drive the effect, providing further support to the idea that numerosity processing exploits discrete inputs.

The Effect of Illusory Contour on the Perception of Symmetry

Computer models are useful tools for evaluating environmental and economic sustainability of proposed dedicated cellulosic grass ecosystems for biofuel production. This study developed an integrated, field scale, and process-based ecosystem model (DRAINMOD-GRASS) for simulating hydrological processes, soil carbon and nitrogen cycling, and plant growth in cropping systems for producing bioenergy grasses in lowland areas. We tested the model using measurements from three replicated switchgrass (Panicum virgatum) plots located in eastern North Carolina, USA. Results showed that the model accurately predicted 5-year (2009–2013) biomass yield. Predicted daily water table depth closely matched field measurements with Nash-Sutcliffe coefficient of 0.86. The model also accurately predicted temporal dynamics of daily soil moisture and temperature with Nash-Sutcliffe coefficients of 0.7 and 0.9, respectively. Predicted seasonal changes in net N mineralization and nitrification rates were comparable to field measurements in 2011 and 2012.

Predictive coding feedback results in perceived illusory contours in a recurrent neural network

Modern feedforward convolutional neural networks (CNNs) can now solve some computer vision tasks at super-human levels. However, these networks only roughly mimic human visual perception. One difference from human vision is that they do not appear to perceive illusory contours (e.g. Kanizsa squares) in the same way humans do. Physiological evidence from visual cortex suggests that the perception of illusory contours could involve feedback connections. Would recurrent feedback neural networks perceive illusory contours like humans? In this work we equip a deep feedforward convolutional network with brain-inspired recurrent dynamics. The network was first pretrained with an unsupervised reconstruction objective on a natural image dataset, to expose it to natural object contour statistics. Then, a classification decision head was added and the model was finetuned on a form discrimination task: squares vs. randomly oriented inducer shapes (no illusory contour). Finally, the model was tested with the unfamiliar “illusory contour” configuration: inducer shapes oriented to form an illusory square. Compared with feedforward baselines, the iterative “predictive coding” feedback resulted in more illusory contours being classified as physical squares. The perception of the illusory contour was measurable in the luminance profile of the image reconstructions produced by the model, demonstrating that the model really “sees” the illusion. Ablation studies revealed that natural image pretraining and feedback error correction are both critical to the perception of the illusion. Finally we validated our conclusions in a deeper network (VGG): adding the same predictive coding feedback dynamics again leads to the perception of illusory contours.

Illusory Planes in Fred Sandback’s Sculpture

Abstract A remarkable feature often noted of artist Fred Sandback’s string constructions: The geometrical forms he created with string have a strong planar feel. Phenomenologically, viewers perceive the spaces between the strings as planes with some substance. The illusion is amodally completed, as in the well-known Kanizsa triangle, by minimal prompts, but in three dimensions. Instead of creating an illusory figure, then, Sandback creates illusory planes. By noting how the constructions are like “impossible” figures, one can see how bottom-up and top-down effects combine to complicate the illusion, and the works become about the construction of space rather than its reification.

Diverse patterns of vulnerability to visual illusions in children with neurodevelopmental disorders.

Research on how children with neurodevelopmental disorders perceive, process, and interpret visual illusions (VIs) has been extensively focused on children with autism spectrum disorder providing controversial findings. In this study, we investigated the patterns of vulnerability to a wide set of VIs comprising 23 standard text book VIs and their variations in a clinical sample of children with neurodevelopmental disorders compared to typically developing children (TD). A total of 176 children, aged between 4.6 and 13.8years old, were distributed into four groups: high-functioning autism (HFA; N = 23), attention-deficit/hyperactivity disorder (ADHD; N = 42), specific learning disorder (SLD; N = 70), and TD (N = 41). Regression models, adjusted for sex, age, and non-verbal IQ, showed that HFA was associated with greater responses accuracy than TD children to the full battery of VIs, to the cognitive illusions, to the distortions, and to both geometrical illusions of size/shape (cognitive distortions) and lightness contrast effects (physical distortions). The susceptibility of ADHD children was found attenuated for illusory contours and greater for paradoxical illusions in comparison with TD children. No significant differences were shown between the SLD group and the TD children. Our findings, which were adjusted for the same duration of visual working memory across groups, showed that there is a potential specific tendency of HFA children to failure of processing visual information in context. Contrarily, children with ADHD showed in general normal global processing such as children diagnosed with SLD.

Interference of Illusory Contour Perception by a Distractor.

The visual system is capable of recognizing objects when object information is widely separated in space, as revealed by the Kanizsa-type illusory contours (ICs). Attentional involvement in perception of ICs is an important topic, and the present study examined whether and how the processing of ICs is interfered with by a distractor. Discrimination between thin and short deformations of an illusory circle was investigated in the absence or presence of a central dynamic patch, with difficulty of discrimination varied in three levels (easy, medium, and hard). Reaction time (RT) was significantly shorter in the absence compared to the presence of the distractor in the easy and medium conditions. Correct rate (CR) was significantly higher in the absence compared to the presence of the distractor in the easy condition, and the magnitude of the difference between CRs of distracted and non-distracted responses significantly reduced as task difficulty increased. These results suggested that perception of ICs is more likely to be vulnerable to distraction when more attentional resources remain available. The present finding supports that attention is engaged in perception of ICs and that distraction of IC processing is associated with perceptual load.