Attention-based U-Net Denoising Network
Diyang Zhu · 2023
Due of its exceptional denoising performance, the denoising technique based on deep convolutional neural networks has attracted a lot of interest. This paper proposes a new Attention-based U-Net to increase the efficacy and practicability of denoising methods. We combine channel attention and spatial attention to better focus on important information in the input image, and apply residual learning and batch normalization to improve network performance. In addition, the quantitative indicators and visual quality evaluations of multiple algorithms on synthetic noise datasets confirm the model's superiority.