Q-DeepSight: Incentivizing Thinking with Images for Image Quality Assessment and Refinement
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866913042291228672 |
|---|---|
| author | Li, Xudong Tan, Jiaxi Zhou, Ziyin Zhong, Yan Huang, Zihao Zheng, Jingyuan Zhang, Yan Zheng, Xiawu Ji, Rongrong |
| author_facet | Li, Xudong Tan, Jiaxi Zhou, Ziyin Zhong, Yan Huang, Zihao Zheng, Jingyuan Zhang, Yan Zheng, Xiawu Ji, Rongrong |
| contents | Image Quality Assessment (IQA) models are increasingly deployed as perceptual critics to guide generative models and image restoration. This role demands not only accurate scores but also actionable, localized feedback. However, current MLLM-based methods adopt a single-look, language-only paradigm, which departs from human evidence-seeking judgment and yields weakly grounded rationales, limiting their reliability for in-the-loop refinement. We propose Q-DeepSight, a think-with-image framework that emulates this human-like process. It performs interleaved Multimodal Chain-of-Thought (iMCoT) with tool-augmented evidence acquisition (e.g., crop-and-zoom) to explicitly determine where quality degrades and why. To train these long iMCoT trajectories via reinforcement learning, we introduce two techniques: Perceptual Curriculum Reward (PCR) to mitigate reward sparsity and Evidence Gradient Filtering (EGF) to improve credit assignment for visually-grounded reasoning. Q-DeepSight achieves state-of-the-art performance across diverse benchmarks, including natural, restored, and AI-generated content. Furthermore, we demonstrate its practical value with Perceptual-in-Generation (PiG), a training-free framework where Q-DeepSight's diagnoses guide iterative image enhancement, effectively closing the loop between assessment and refinement. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_16858 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Q-DeepSight: Incentivizing Thinking with Images for Image Quality Assessment and Refinement Li, Xudong Tan, Jiaxi Zhou, Ziyin Zhong, Yan Huang, Zihao Zheng, Jingyuan Zhang, Yan Zheng, Xiawu Ji, Rongrong Computer Vision and Pattern Recognition Image Quality Assessment (IQA) models are increasingly deployed as perceptual critics to guide generative models and image restoration. This role demands not only accurate scores but also actionable, localized feedback. However, current MLLM-based methods adopt a single-look, language-only paradigm, which departs from human evidence-seeking judgment and yields weakly grounded rationales, limiting their reliability for in-the-loop refinement. We propose Q-DeepSight, a think-with-image framework that emulates this human-like process. It performs interleaved Multimodal Chain-of-Thought (iMCoT) with tool-augmented evidence acquisition (e.g., crop-and-zoom) to explicitly determine where quality degrades and why. To train these long iMCoT trajectories via reinforcement learning, we introduce two techniques: Perceptual Curriculum Reward (PCR) to mitigate reward sparsity and Evidence Gradient Filtering (EGF) to improve credit assignment for visually-grounded reasoning. Q-DeepSight achieves state-of-the-art performance across diverse benchmarks, including natural, restored, and AI-generated content. Furthermore, we demonstrate its practical value with Perceptual-in-Generation (PiG), a training-free framework where Q-DeepSight's diagnoses guide iterative image enhancement, effectively closing the loop between assessment and refinement. |
| title | Q-DeepSight: Incentivizing Thinking with Images for Image Quality Assessment and Refinement |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2604.16858 |