Miao, Y., Deng, J. and Han, J. (Accepted: 2024) WaveFace: authentic face restoration with efficient frequency recovery. In: Proceedings of The IEEE / CVF Computer Vision and Pattern Recognition Conference. IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR), 21 Jun 2024, Seattle WA, USA. Institute of Electrical and Electronics Engineers (IEEE) . (In Press)
Abstract
Although diffusion models are rising as a powerful solution for blind face restoration, they are criticized for two problems: 1) slow training and inference speed, and 2) failure in preserving identity and recovering fine-grained facial details. In this work, we propose WaveFace to solve the problems in the frequency domain, where low- and highfrequency components decomposed by wavelet transformation are considered individually to maximize authenticity as well as efficiency. The diffusion model is applied to recover the low-frequency component only, which presents general information of the original image but 1/16 in size. To preserve the original identity, the generation is conditioned on the low-frequency component of low-quality images at each denoising step. Meanwhile, high-frequency components at multiple decomposition levels are handled by a unified network, which recovers complex facial details in a single step. Evaluations on four benchmark datasets show that: 1) WaveFace outperforms state-of-the-art methods in authenticity, especially in terms of identity preservation, and 2) authentic images are restored with the efficiency 10× faster than existing diffusion model-based BFR methods.
Metadata
Item Type: | Proceedings Paper |
---|---|
Authors/Creators: |
|
Copyright, Publisher and Additional Information: | © 2024 The authors. |
Dates: |
|
Institution: | The University of Sheffield |
Academic Units: | The University of Sheffield > Faculty of Engineering (Sheffield) > Department of Computer Science (Sheffield) |
Depositing User: | Symplectic Sheffield |
Date Deposited: | 12 Apr 2024 13:51 |
Last Modified: | 12 Apr 2024 13:51 |
Status: | In Press |
Publisher: | Institute of Electrical and Electronics Engineers (IEEE) |
Refereed: | Yes |
Related URLs: |
Download
Filename: CVPR_2024_wavelet.pdf