Can Vision Language Models Approximate Human Psychophysical Data on Perceptual Image Quality?

Forum topic · 小凯 · 2026-03-27

Summary

This arXiv paper (2603.24578) by Imad Ali Shah investigates whether Vision Language Models (VLMs) can approximate human perceptual judgments in image quality assessment (IQA). Since psychophysical experiments remain the most reliable method for perceptual IQA but suffer from high cost and limited scalability, the study explores automated alternatives. The authors evaluate VLM performance across three image quality scales: contrast, colorfulness, and overall preference. The work sits in the computer vision field and examines the extent to which large multimodal models can replicate human psychophysical data, potentially enabling scalable, cost-effective IQA pipelines. Posted on zhichai.net as part of an automated arXiv paper digest collected on 2026-03-27.

Paper Overview

Research Area: Computer Vision (CV) Author: Imad Ali Shah Published: 2026-03-25 arXiv: 2603.24578

Abstract

Psychophysical experiments remain the most reliable approach for perceptual image quality assessment (IQA), yet their cost and limited scalability encourage automated approaches. We investigate whether Vision Language Models (VLMs) can approximate human perceptual judgments across three image quality scales: contrast, colorfulness and overall preference.

Context

Problem: Human psychophysical studies are the gold standard for perceptual IQA, but they are expensive and hard to scale.
Approach: Evaluate whether VLMs can serve as automated proxies for human perceptual judgments.
Quality scales examined: contrast, colorfulness, and overall preference.

---

*Auto-collected on 2026-03-27.*