SR-Ground: Image Quality Grounding for Super-Resolved Content

2026-05-20

AI Summary

This work addresses the challenge of fine-grained identification and interpretable analysis of visual artifacts in super-resolved images, which existing image quality assessment (IQA) methods struggle to achieve. To this end, the authors introduce SR-Ground, a large-scale dataset comprising 63,000 images generated by diverse state-of-the-art super-resolution models, accompanied by the first pixel-level annotations for six distinct artifact types. Annotation quality is ensured through a combination of automated segmentation and validation by 1,062 crowd-sourced participants. Leveraging this dataset, the study develops a localization-aware IQA model training framework and a fine-tuning pipeline for artifact suppression. Experiments demonstrate that the proposed approach significantly enhances downstream task performance and effectively reduces visible artifacts in super-resolution outputs.

Abstract

Super-Resolution (SR) has advanced rapidly in recent years, with diffusion-based models achieving unprecedented fidelity at the cost of introducing new types of visual artifacts. While existing Image Quality Assessment (IQA) methods provide holistic quality scores, they lack interpretability and fail to distinguish between different artifact types arising from modern SR approaches. To address this gap, we introduce SR-Ground, a large-scale dataset specifically designed for fine-grained artifact segmentation in super-resolved images. The dataset comprises images processed by a diverse set of state-of-the-art SR models, with pixel-level annotations for multiple artifact categories. We conduct a large-scale crowdsourcing study involving 1,062 participants to validate and refine automatically generated segmentations, resulting in a high-quality dataset of 63,000 images spanning 6 distinct artifact types. We demonstrate that training IQA models with grounding capabilities on SR-Ground significantly improves performance on downstream tasks. Furthermore, we introduce a fine-tuning pipeline that leverages our grounding model to reduce perceptible artifacts in SR outputs, showcasing the practical utility of our dataset.

Problem

• Super-Resolution
• Image Quality Assessment
• Visual Artifacts
• Artifact Segmentation
• Interpretability

Innovation

• artifact segmentation
• super-resolution
• image quality assessment
• pixel-level annotation
• diffusion-based models