Authors: Conghan Yue, Zhengwei Peng, Junlong Ma, Shiyan Du, Pengxu Wei, Dongyu Zhang
Published on: December 16, 2023
Impact Score: 6.6
Arxiv code: Arxiv:2312.10299
Summary
- What is new: Introduction of the Generalized Ornstein-Uhlenbeck Bridge (GOUB) model for image restoration, and demonstration of its empirical optimality across bridge models.
- Why this is important: The need for improved mapping from low-quality to high-quality images in image restoration tasks.
- What the research proposes: Utilizing the GOUB model that leverages a natural mean-reverting process and an advanced statistical transform (Doob’s h-transform) to enhance image quality effectively.
- Results: The GOUB and Mean-ODE models demonstrated state-of-the-art results in tasks like inpainting, deraining, and super-resolution.
Technical Details
Technological frameworks used: nan
Models used: Generalized Ornstein-Uhlenbeck Bridge (GOUB), Mean-ODE model
Data used: Image datasets for inpainting, deraining, and super-resolution tasks
Potential Impact
Photography technology companies, mobile and desktop image editing software markets, and digital media production industries
Want to implement this idea in a business?
We have generated a startup concept here: Clearify.
Leave a Reply