SEAGULL: No-reference Image Quality Assessment for Regions of Interest via Vision-Language Instruction Tuning

TL;DR

SEAGULL is a novel vision-language based network for fine-grained, region-specific image quality assessment that leverages large models, specialized datasets, and mask-based features to improve ROI quality analysis.

Contribution

The paper introduces SEAGULL, a new model combining vision-language guidance and mask-based features for ROI IQA, along with two new datasets for training and evaluation.

Findings

SEAGULL outperforms existing methods in ROI quality assessment.

Pre-training on SEAGULL-100w enhances regional quality perception.

Fine-tuning on SEAGULL-3k improves real-world distortion detection.

Abstract

Existing Image Quality Assessment (IQA) methods achieve remarkable success in analyzing quality for overall image, but few works explore quality analysis for Regions of Interest (ROIs). The quality analysis of ROIs can provide fine-grained guidance for image quality improvement and is crucial for scenarios focusing on region-level quality. This paper proposes a novel network, SEAGULL, which can SEe and Assess ROIs quality with GUidance from a Large vision-Language model. SEAGULL incorporates a vision-language model (VLM), masks generated by Segment Anything Model (SAM) to specify ROIs, and a meticulously designed Mask-based Feature Extractor (MFE) to extract global and local tokens for specified ROIs, enabling accurate fine-grained IQA for ROIs. Moreover, this paper constructs two ROI-based IQA datasets, SEAGULL-100w and SEAGULL-3k, for training and evaluating ROI-based IQA. SEAGULL-100w comprises about 100w synthetic distortion images with 33 million ROIs for pre-training to improve the model's ability of regional quality perception, and SEAGULL-3k contains about 3k authentic distortion ROIs to enhance the model's ability to perceive real world distortions. After pre-training on SEAGULL-100w and fine-tuning on SEAGULL-3k, SEAGULL shows remarkable performance on fine-grained ROI quality assessment. Code and datasets are publicly available at the https://github.com/chencn2020/Seagull.