Summary of Zero-shot Video Restoration and Enhancement Using Pre-trained Image Diffusion Model, by Cong Cao et al.
Zero-Shot Video Restoration and Enhancement Using Pre-Trained Image Diffusion Model
by Cong Cao, Huanjing Yue, Xin Liu, Jingyu Yang
First submitted to arxiv on: 2 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes a framework for zero-shot video restoration and enhancement, building upon the success of diffusion-based image models. It addresses the issue of temporal flickering artifacts when directly applying these models to videos. The proposed method replaces spatial self-attention with short-long-range (SLR) temporal attention, allowing the model to leverage frame correlations. Additionally, it introduces temporal consistency guidance, spatial-temporal noise sharing, and early stopping sampling strategy to improve temporally consistent sampling. This plug-and-play module can be integrated into existing image restoration or enhancement methods, enhancing their performance. Experimental results demonstrate its superiority. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a way for videos to get restored and improved without needing any extra training data. Currently, this type of technology works great for still images, but when used on videos, it makes them look like they’re flickering. The researchers fix this problem by changing the way the model looks at different parts of the video. They also add some new techniques to make sure the restored video looks consistent and natural. This method can be added to existing image restoration technology to make it even better. |
Keywords
* Artificial intelligence * Attention * Diffusion * Early stopping * Self attention * Zero shot
