Gated Spatial-Frequency Fusion Enhanced Image Denoising

Image denoising is a fundamental challenge in low-level vision, crucial for applications such as medical imaging and remote sensing. Traditional methods often rely on single-domain modeling, limiting their ability to exploit complementary spatial and frequency characteristics. This paper introduces SFUNet, a dual-domain denoising framework that integrates spatial Transformer and wavelet convolution branches through an adaptive gated fusion mechanism. Extensive experiments on synthetic and real-world datasets demonstrate that SFUNet achieves state-of-the-art denoising quality, maintaining stable generalization under varying noise distributions. Our design provides a generic paradigm for cross-domain representation learning, shedding light on integrating spatial-frequency synergy in low-level vision tasks. The code is available at https://github.com/ywxk111/SFUNet