I am a Ph.D. student at the Computer Vision Lab, Seoul National University, advised by Prof. Kyoung Mu Lee.
My research focuses on application-oriented image restoration, with a primary emphasis on task-driven image restoration, which develops restoration models to improve the performance of downstream high-level vision tasks. Motivated by challenges in practical deployment, my previous work also includes real-world image restoration, which aims to recover high-quality images from realistically degraded inputs.
Recently, I have been particularly interested in leveraging foundation models, such as diffusion models and vision-language models, as powerful priors and tools for image restoration.
You can download my CV here 📄.
🔥 News
- 2026.07: 📄 New preprint is available on arXiv!
- 2026.07: 🎉🎉 One paper got accepted to ECCV 2026!
- 2025.06: 🎉🎉 One paper got accepted to ICCV 2025!
- 2025.06: 🏆 Recognized as Outstanding Reviewer at CVPR 2025!
📝 Publications
Noise-Free One-Step LoRA for Task-Driven Image Restoration with Diffusion Priors (First author) Jaeha Kim and Kyoung Mu Lee
- Proposing a deterministic one-step adaptation for stable task-driven restoration.
- Code
Learning to Corrupt for Better Restoration (Author) Joonkyu Park, Wooseok Lee, Jaeha Kim, Sehoon Kim, Bokyeung Lee, and Kyoung Mu Lee
- Proposing input-aware corruption modeling to improve perceptual quality in diffusion-based image restoration.
Exploiting Diffusion Prior for Task-driven Image Restoration (First author) Jaeha Kim, Junghun Oh, and Kyoung Mu Lee
- Task-driven diffusion-based image restoration for improved high-level vision tasks.
- Code
Beyond Image Super-Resolution for Image Recognition with Task-Driven Perceptual Loss (First author) Jaeha Kim, Junghun Oh, and Kyoung Mu Lee
- A task-driven super-resolution framework for image recognition.
- Code
Recovering 3D Hand Mesh Sequence from a Single Blurry Image: A New Dataset and Temporal Unfolding (Co-first author) Yeoung Uk Oh*, JoonKyu Park*, Jaeha Kim*, Gyeongsik Moon, Kyoung Mu Lee
- A new practical benchmark for 3D hand pose estimation under blur.
- Code
Toward Real-World Super-Resolution via Adaptive Downsampling Models (Co-first author) Sanghyun Son*, Jaeha Kim*, Wei-Sheng Lai, Ming-Hsuan Yang, Kyoung Mu Lee
🎖 Honors and Awards
- 2025.06 Outstanding Reviewer (Top 5%), CVPR 2025.
- 2019.02 National Scholarship (Undergraduate), Korea Student Aid Foundation.
📖 Educations
- 2019.09 - present, Seoul National University, Seoul, Korea.
- 2015.03 - 2019.02, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea.
- 2013.03 - 2015.02, Incheon Science High School, Incheon, Korea.