Prompt-Driven Color Accessibility Evaluation in Diffusion-based Image Generation Models

2026-03-10

AI Summary

This study addresses the lack of color accessibility for color vision deficiency (CVD) users in current diffusion models, which struggle to reliably respond to accessibility-oriented prompts. The authors systematically evaluate mainstream diffusion models’ ability to generate CVD-friendly images when guided by such prompts and introduce a novel metric, CVDLoss, which quantifies structural detail preservation through image gradient differences to assess the effectiveness of accessibility enhancements. By integrating off-the-shelf CVD simulation tools with prompt engineering, experiments reveal that diffusion models exhibit inconsistent responsiveness to accessibility prompts. However, CVDLoss demonstrates strong alignment with established daltonization methods and effectively evaluates the perceptual quality of generated images, offering a reliable tool for optimizing color accessibility in generative models.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Humans and AI: AI for Accessibility
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Generative models are increasingly integrated into creative workflows. While text-to-image generation excels in visual quality and diversity, color accessibility for users with Color Vision Deficiencies (CVD) remains largely unexplored. Our work systematically evaluates color accessibility in images generated by a common pretrained diffusion model, prompted to improve accessibility across diverse categories. We quantify performance using established, off-the-shelf CVD simulation methods and introduce "CVDLoss", a new metric measuring differences in image gradients indicative of structural detail. We validate CVDLoss against a commonly used daltonization method, demonstrating its sensitivity to color accessibility modifications. Applying CVDLoss to model outputs reveals that existing diffusion models struggle to reliably respond to accessibility-focused prompts. Consequently, our study establishes CVDLoss as a valuable evaluation tool for accessibility-aware image generation and post-processing, offering insights into current generative models' limitations in addressing color accessibility.

Problem

• color accessibility
• Color Vision Deficiencies
• diffusion models
• text-to-image generation
• prompt-driven generation

Innovation

• CVDLoss
• color accessibility
• diffusion models
• Color Vision Deficiency
• prompt-driven evaluation