Diffusion based image super resolution quality assessment dataset released
DISRQAD: Diffusion Image Super-Resolution Quality Assessment Dataset and Benchmark
Computer Vision and Pattern Recognition
Summary
Images made clearer by diffusion-based super-resolution can add realistic details not in the original low-resolution picture, making it tricky to judge their quality. The authors created DISRQAD, a large set of human ratings for different super-resolution methods, focusing on diffusion-based ones, to help measure how good these enhanced images really look. They tested many existing quality metrics and found these often don’t agree well with human opinions on diffusion-enhanced images. Their dataset helps identify gaps in current quality checks and supports building new tools that better recognize artifacts specific to diffusion methods.
What this means in practice
- •For image processing engineers: Use DISRQAD to benchmark and improve quality assessment tools for diffusion-based super-resolution images under various low-resolution conditions.
- •For digital media product teams: Evaluate and choose super-resolution methods that better match user perceptions for image enhancement features in apps or services.$Commercial implications: Enables development of user-aligned image enhancement products by providing human-rated quality data specific to diffusion-based methods.
Authors
Nikita Kukuzei, Artem Borisov, Evgeney Bogatyrev, Khaled Abud, Egor Chistov, Dmitriy Vatolin
Abstract
Diffusion-based image super-resolution (SR) can create visually plausible detail that is not supported by the low-resolution input. We introduce DISRQAD, a subjective-quality dataset and diagnostic benchmark for this setting. It contains mean opinion scores (MOS) for 14,000 SR outputs from ten diffusion and four non-diffusion methods, spanning four low-resolution degradation conditions and x2/x4 upscaling. We evaluate 51 standard full-reference and no-reference metric configurations and 11 adapted variants. Agreement with MOS is substantially weaker on diffusion outputs: the strongest standard no-reference baseline reaches 0.431 SRCC on diffusion SR versus 0.813 on non-diffusion SR. As a case study in benchmark use, a pruned and distilled Q-ReAlign-mini student reaches 0.496 SRCC on diffusion SR. DISRQAD measures perceived output quality, not faithfulness to the input; it enables analysis of metric behavior across generator families and input conditions. Our findings reveal a substantial gap in the assessment of diffusion-based SR and provide a basis for developing quality models sensitive to diffusion-specific artifacts.