VOSR: A Vision-Only Generative Model for Image Super-Resolution
Paper • 2604.03225 • Published
Image/Video Restoration, Enhancement and Quality Assessment
This collection hosts VCLab's leading research in visual restoration, enhancement, and quality assessment, especially in real-world scenarios.
Note [CVPR 2026] Vision-only generative SR trained without text-image pairs. | Code: https://github.com/cswry/VOSR
GDPO-SR: Group Direct Preference Optimization for One-Step Generative Image Super-Resolution
Paper • 2603.16769 • PublishedNote [CVPR 2026] Group DPO for one-step generative SR. | Code: https://github.com/Joyies/GDPO
VisualQuality-R1: Reasoning-Induced Image Quality Assessment via Reinforcement Learning to Rank
Paper • 2505.14460 • Published • 33Note [NeurIPS 2025 Spotlight] Reasoning-induced IQA via RL2R. | Code: https://github.com/TianheWu/VisualQuality-R1 | HF: https://huggingface.co/TianheWu/VisualQuality-R1
DP^2O-SR: Direct Perceptual Preference Optimization for Real-World Image Super-Resolution
Paper • 2510.18851 • PublishedNote [NeurIPS 2025] Direct perceptual preference optimization for real-world SR. | Code: https://github.com/cswry/DP2O-SR
One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution
Paper • 2506.15591 • PublishedNote [NeurIPS 2025] One-step diffusion VSR with temporal consistency. | Code: https://github.com/yjsunnn/DLoRAL
Fine-structure Preserved Real-world Image Super-resolution via Transfer VAE Training
Paper • 2507.20291 • PublishedNote [ICCV 2025 Highlight] Transfer VAE training preserves fine structures. | Code: https://github.com/Joyies/TVT
Toward Generalized Image Quality Assessment: Relaxing the Perfect Reference Quality Assumption
Paper • 2503.11221 • Published • 1Note [CVPR 2025] Generalized IQA without perfect reference. | Code: https://tianhewu.github.io/A-FINE-page.github.io/
Pixel-level and Semantic-level Adjustable Super-resolution: A Dual-LoRA Approach
Paper • 2412.03017 • PublishedNote [CVPR 2025] Dual-LoRA for pixel+semantic adjustable SR. | Code: https://github.com/csslc/PiSA-SR | HF: https://huggingface.co/csslc/PiSA-SR
One-Step Effective Diffusion Network for Real-World Image Super-Resolution
Paper • 2406.08177 • PublishedNote [NeurIPS 2024] One-step diffusion for Real-ISR. | Code: https://github.com/cswry/OSEDiff | HF: https://huggingface.co/cswry/OSEDiff
Pixel-Aware Stable Diffusion for Realistic Image Super-resolution and Personalized Stylization
Paper • 2308.14469 • Published • 7Note [ECCV 2024] Pixel-aware SD for Real-ISR & stylization. | Code: https://github.com/yangxy/PASD
Motion-Guided Latent Diffusion for Temporally Consistent Real-world Video Super-resolution
Paper • 2312.00853 • PublishedNote [ECCV 2024] Motion-guided latent diffusion for VSR. | Code: https://github.com/IanYeung/MGLD-VSR
SeeSR: Towards Semantics-Aware Real-World Image Super-Resolution
Paper • 2311.16518 • PublishedNote [CVPR 2024] Semantics-aware Real-ISR with SD. | Code: https://github.com/cswry/SeeSR | HF: https://huggingface.co/Iceclear/SeeSR
Details or Artifacts: A Locally Discriminative Learning Approach to Realistic Image Super-Resolution
Paper • 2203.09195 • PublishedNote [CVPR 2022 Oral] Locally discriminative learning for Real-ISR. | Code: https://github.com/csjliang/LDL
A Text Attention Network for Spatial Deformation Robust Scene Text Image Super-resolution
Paper • 2203.09388 • PublishedNote [CVPR 2022] Text-attention for scene-text SR. | Code: https://github.com/mjq11302010044/TATT
Dual Adversarial Network: Toward Real-world Noise Removal and Noise Generation
Paper • 2007.05946 • PublishedNote [ECCV 2020] Dual adversarial net for real denoising. | Code: https://github.com/zsyOAOA/DANet
Plug-and-Play Image Restoration with Deep Denoiser Prior
Paper • 2008.13751 • PublishedNote [TPAMI 2021] Plug-and-play IR with deep denoiser. | Code: https://github.com/cszn/DPIR
Text Prior Guided Scene Text Image Super-resolution
Paper • 2106.15368 • PublishedNote [TIP 2023] Text-prior guided scene-text SR. | Code: https://github.com/mjq11302010044/TPGSR
NSARM: Next-Scale Autoregressive Modeling for Robust Real-World Image Super-Resolution
Paper • 2510.00820 • PublishedNote [arXiv 2025] Next-scale AR modeling for Real-ISR. | Code: https://github.com/Xiangtaokong/NSARM
-
InstructRestore: Region-Customized Image Restoration with Human Instructions
Paper • 2503.24357 • Published -
Perceive, Understand and Restore: Real-World Image Super-Resolution with Autoregressive Multimodal Generative Models
Paper • 2503.11073 • Published • 1 -
Generalized and Efficient 2D Gaussian Splatting for Arbitrary-scale Super-Resolution
Paper • 2501.06838 • Published -
Improving the Stability of Diffusion Models for Content Consistent Super-Resolution
Paper • 2401.00877 • Published