Problem
Research questions and friction points this paper is trying to address.
diffusion-based image super-resolution
quality assessment
subjective quality dataset
image quality metrics
benchmark
DISRQAD: Diffusion Image Super-Resolution Quality Assessment Dataset and Benchmark
🤖 AI Summary
Diffusion-based super-resolution models frequently generate hallucinated details, rendering existing quality assessment metrics ineffective. This study addresses this limitation by constructing DISRQAD, the first diagnostic benchmark tailored for this task, comprising 140,000 subjective scores to systematically evaluate 51 quality metrics. Furthermore, by integrating knowledge distillation with model pruning, this work proposes Q-ReAlign-mini, a lightweight assessment model. Our analysis reveals a substantial misalignment between standard metrics and human perception, with the best-performing no-reference metric achieving a Spearman’s rank correlation coefficient (SRCC) of only 0.431. In contrast, Q-ReAlign-mini improves the SRCC to 0.496, effectively bridging the evaluation gap in diffusion-based super-resolution scenarios.
📝 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.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Diffusion-based Super-Resolution
Quality Assessment Dataset
Diagnostic Benchmark
Image Quality Metrics
Knowledge Distillation
🔎 Similar Papers
No similar papers found.
