DenoiseNet: towards advanced denoising with transformer network
Liang Shi, Dunwei Han, Guifang Luo, Qing Chen, Lin Tang, Zhongping Zhai, Hui Xu · 2025
Image denoising is a fundamental challenge in image restoration. In this paper, we propose DenoiseNet, a Transformer-based network designed to effectively address this problem. DenoiseNet introduces two key components: the BGFN module, which applies dual filtering to reduce image disturbances and preserve original content, the SOSA module, which computes attention across both spatial dimensions of the image to optimize denoising at multiple levels. Experimental results demonstrate that DenoiseNet outperforms other methods on public datasets, achieving significant improvements in denoising quality.