TAPE: Task-Agnostic Prior Embedding for Image Restoration

• Published in: Computer Vision - ECCV 2022 Vol. 13678; pp. 447 - 464
• Main Authors: Liu, Lin, Xie, Lingxi, Zhang, Xiaopeng, Yuan, Shanxin, Chen, Xiangyu, Zhou, Wengang, Li, Houqiang, Tian, Qi
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer 2022; Springer Nature Switzerland
• Series: Lecture Notes in Computer Science
• ISBN: 9783031197963; 3031197968
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-031-19797-0_26

Learning a generalized prior for natural image restoration is an important yet challenging task. Early methods mostly involved handcrafted priors including normalized sparsity, ℓ0 $$\ell _0$$ gradients, dark channel priors, etc. Recently, deep neural networks have been used to learn various image priors but do not guarantee to generalize. In this paper, we propose a novel approach that embeds a task-agnostic prior into a transformer. Our approach, named Task-Agnostic Prior Embedding (TAPE), consists of two stages, namely, task-agnostic pre-training and task-specific fine-tuning, where the first stage embeds prior knowledge about natural images into the transformer and the second stage extracts the knowledge to assist downstream image restoration. Experiments on various types of degradation validate the effectiveness of TAPE. The image restoration performance in terms of PSNR is improved by as much as 1.45 dB and even outperforms task-specific algorithms. More importantly, TAPE shows the ability of disentangling generalized image priors from degraded images, which enjoys favorable transfer ability to unknown downstream tasks.