Attention Methods for Image Denoising: A Survey

Image denoising is an important branch of image processing, which is developed from traditional mechina learning methods to deep learning methods. Over the past few years, the focus on attention mechanisms has grown, owing to their substantial impact on various image processing applications. After the Transformer architecture was proposed, more attention mechanism-based research and improvements were proposed. However, there are few researches on image denoising based on different attention techniques. Hence, this paper delivers an exhaustive examination and overview of the utilization of attention mechanisms for image noise reduction over the past few years. We have presented three prevalent attention strategies: spatial attention, channel attention, and self-attention, and evaluated the efficacy of attention-driven image denoising techniques. Finally, this paper discusses the problems and challenges in current research and provides some inspiration for future research directions.