GFFNet: An Efficient Image Denoising Network with Group Feature Fusion

Image denoising is a critical pre-processing step for a wide range of image processing and computer vision applications, where the primary goal is to remove noise interference from corrupted images while preserving the essential features of the image. Although recent research has made significant progress in images denoising using deep learning methods, problems such as loss of detail, difficulty in recovering edge textures, and low image processing performance still persist. To tackle these issues, we develope an effective network architecture. This study introduces a Group Feature Fusion (GFF) module, which leverages image feature grouping and fusion techniques to enhance the representation capacity and computational efficiency of features in our network. Additionally, this artical introduce a Cross-Information Integration (CII) Module to enhance the network’s ability to utilize input data features by integrating low-level and high-level channel information. Finally, the network was enhanced in its effectiveness for edge texture restoration by optimizing it with the PSNR loss function in conjunction with a novel edge loss function. This architecture achieves significant performance improvements on nine benchmark test datasets for image denoising tasks. Extensive experiments have demonstrated the efficiency and superior performance of this architecture.