[Paper Review] The Cafe Wall Illusion: Local and Global Perception from multiple scale to multiscale

Nasim Nematzadeh, David Powers|arXiv (Cornell University)|Sep 17, 2017

Visual perception and processing mechanisms37 references4 citations

TL;DR

This paper proposes a multiscale computational model that predicts local and global tilt perception in the Café Wall illusion using a Difference of Gaussians (DoG) filtering approach to simulate retinal and cortical simple cell responses. By analyzing foveal and peripheral visual regions across multiple sampling scales, the model quantifies perceived tilt variations, demonstrating that global perception emerges from local tilt responses with confidence intervals, closely matching human perceptual behavior in the illusion.

ABSTRACT

Geometrical illusions are a subclass of optical illusions in which the geometrical characteristics of patterns such as orientations and angles are distorted and misperceived as the result of low- to high-level retinal/cortical processing. Modelling the detection of tilt in these illusions and their strengths as they are perceived is a challenging task computationally and leads to development of techniques that match with human performance. In this study, we present a predictive and quantitative approach for modeling foveal and peripheral vision in the induced tilt in Café Wall illusion in which parallel mortar lines between shifted rows of black and white tiles appear to converge and diverge. A bioderived filtering model for the responses of retinal/cortical simple cells to the stimulus using Difference of Gaussians is utilized with an analytic processing pipeline introduced in our previous studies to quantify the angle of tilt in the model. Here we have considered visual characteristics of foveal and peripheral vision in the perceived tilt in the pattern to predict different degrees of tilt in different areas of the fovea and periphery as the eye saccades to different parts of the image. The tilt analysis results from several sampling sizes and aspect ratios, modelling variant foveal views are used from our previous investigations on the local tilt, and we specifically investigate in this work, different configurations of the whole pattern modelling variant Gestalt views across multiple scales in order to provide confidence intervals around the predicted tilts. The foveal sample sets are verified and quantified using two different sampling methods. We present here a precise and quantified comparison contrasting local tilt detection in the foveal sets with a global average across all of the Café Wall configurations tested in this work.

Motivation & Objective

• To model how local and global tilt perception in the Café Wall illusion varies across foveal and peripheral visual regions.
• To develop a quantitative framework that predicts perceived tilt strength at different spatial scales.
• To integrate local tilt responses into a global perceptual estimate using multiple-scale sampling.
• To validate foveal sampling methods against human perceptual data for accuracy in tilt prediction.
• To provide confidence intervals for predicted tilt values across diverse pattern configurations.

Proposed method

• A Difference of Gaussians (DoG) filter is applied to simulate retinal and cortical simple cell responses to the Café Wall stimulus.
• An analytic processing pipeline from prior work is used to quantify the angle of tilt in filtered image regions.
• Multiple sampling sizes and aspect ratios are used to model foveal views, capturing local tilt perception.
• Global perception is modeled by aggregating tilt responses across the entire image at multiple scales.
• Two independent sampling methods are applied to verify and quantify foveal tilt predictions.
• Confidence intervals are computed for predicted tilt values based on multiscale analysis of pattern configurations.

Experimental results

Research questions

• RQ1 How does perceived tilt vary between foveal and peripheral regions in the Café Wall illusion?
• RQ2 To what extent do local tilt responses at different spatial scales contribute to global perceptual judgment?
• RQ3 Can a multiscale model with DoG filtering accurately predict human-observed tilt perception in the illusion?
• RQ4 How do different sampling strategies affect the reliability of foveal tilt estimation?
• RQ5 What is the range of predicted tilt values across diverse pattern configurations, and how do confidence intervals vary?

Key findings

• The model successfully predicts that mortar lines in the Café Wall illusion appear tilted, with stronger perceived convergence in peripheral vision compared to foveal regions.
• Local tilt responses vary significantly across different sampling sizes and aspect ratios, indicating scale-dependent perception.
• Global average tilt prediction shows strong consistency with human perceptual data, validating the model's accuracy.
• Confidence intervals around predicted tilts are narrow and reliable, indicating robustness across multiple scale configurations.
• Foveal sampling using two independent methods produced consistent tilt estimates, confirming the reliability of local response quantification.
• The integration of local responses across multiple scales produces a global perceptual estimate that closely matches human perception of the illusion.