Abstract:Perception is often viewed as a process that transforms physical variables, external to an observer, into internal psychological variables. Such a process can be modeled by a function coined perceptual scale. The perceptual scale can be deduced from psychophysical measurements that consist in comparing the relative differences between stimuli (i.e. difference scaling experiments). However, this approach is often overlooked by the modeling and experimentation communities. Here, we demonstrate the value of measuring the perceptual scale of classical (spatial frequency, orientation) and less classical physical variables (interpolation between textures) by embedding it in recent probabilistic modeling of perception. First, we show that the assumption that an observer has an internal representation of univariate parameters such as spatial frequency or orientation while stimuli are high-dimensional does not lead to contradictory predictions when following the theoretical framework. Second, we show that the measured perceptual scale corresponds to the transduction function hypothesized in this framework. In particular, we demonstrate that it is related to the Fisher information of the generative model that underlies perception and we test the predictions given by the generative model of different stimuli in a set a of difference scaling experiments. Our main conclusion is that the perceptual scale is mostly driven by the stimulus power spectrum. Finally, we propose that this measure of perceptual scale is a way to push further the notion of perceptual distances by estimating the perceptual geometry of images i.e. the path between images instead of simply the distance between those.

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve The paper attempts to address the following key issues: 1. **Measurement and Modeling of Perceptual Scales**: - Researchers demonstrate the importance of quantifying how physical variables are perceived by humans through perceptual scales. Perceptual scales can be inferred by comparing the relative differences between stimuli (i.e., difference scale experiments). However, this method is often overlooked in the modeling and experimental community. The authors of this paper show the value of measuring perceptual scales for classical physical variables (such as spatial frequency and orientation) as well as less common physical variables (such as texture interpolation) and embedding them into recent probabilistic perceptual modeling. 2. **Theoretical Framework and Empirical Validation**: - The researchers hypothesize that observers have an internal representation of univariate parameters (such as spatial frequency or orientation), even if the stimuli are high-dimensional, and this hypothesis does not lead to contradictory predictions within the theoretical framework. Additionally, they demonstrate that the measured perceptual scales are consistent with the transduction functions assumed in this framework. Specifically, it is related to the Fisher information of the generative model and tests the predictions of different stimuli in a set of difference scale experiments. The main conclusion of the study is that perceptual scales are primarily driven by the energy spectrum of the stimuli. 3. **Further Exploration of Perceptual Distance**: - The authors propose a new method to further expand the concept of perceptual distance, by estimating the perceptual geometry of images to measure the path between images, rather than simply measuring the distance between them. Specifically, they test the predictions given by the Fisher information of Gaussian vectors and processes through a series of experiments (code and data are available in the paper's appendix), involving random stimuli characterized by their power spectrum or higher-order statistical features (captured by VGG-19). 4. **Quantification of Perceptual Consistency**: - Finally, the authors propose the "Area Matching Score" (AMS) to quantify the perceptual consistency between generative image models and human vision. This provides a clear method for evaluating the consistency between generative image models and human visual perception.

Perceptual Scales Predicted by Fisher Information Metrics

A unified framework for perceived magnitude and dicriminability of sensory stimuli

Perceptual Experience As a Bridge Between the Retina and a Bicoded Cognitive Map.

Perceptual Scale Expansion: an Efficient Angular Coding Strategy for Locomotor Space.

A Fisher Information Theory of Aesthetic Preference for Complexity

Image Statistics Predict the Sensitivity of Perceptual Quality Metrics

Spatial frequency channels mediate a visual metric for human spatial perception

Perceptual Scale Space and Its Applications.

Evaluating Perceptual Distance Models by Fitting Binomial Distributions to Two-Alternative Forced Choice Data

Psychophysical scaling reveals a unified theory of visual memory strength

Quantifying Signal-to-Noise Ratio in Neural Latent Trajectories via Fisher Information

Informing Machine Perception With Psychophysics

Derivation of Human Chromatic Discrimination Ability from an Information-Theoretical Notion of Distance in Color Space

Perceptual Scale-Space and Its Application

Perceptual Scale Space and Its Applications

Measuring Stimulus Information Transfer Between Neural Populations through the Communication Subspace

Approximations to the Fisher Information Metric of Deep Generative Models for Out-Of-Distribution Detection

Perceptual Score: What Data Modalities Does Your Model Perceive?

Likelihood method and Fisher information in construction of physical models

Low-level features predict perceived similarity for naturalistic images