Two-stage Progressive Residual Dense Attention Network for Image Denoising

• DOI: 10.48550/arxiv.2401.02831
• Publication Date: 2024-01-01
• Authors: Wencong Wu; An Ge; Guannan Lv; Yuelong Xia; Yungang Zhang; Wen Xiong
• Keywords: Benchmark (surveying); Feature (linguistics); Code (set theory)

Deep convolutional neural networks (CNNs) for image denoising can effectively exploit rich hierarchical features and have achieved great success. However, many deep CNN-based models equally utilize the of noisy images without paying attention to more important useful features, leading relatively low performance. To address issue, we design a new Two-stage Progressive Residual Dense Attention Network (TSP-RDANet) denoising, which divides whole process into two sub-tasks remove noise progressively. Two different mechanism-based are designed sequential sub-tasks: residual dense module (RDAM) is first stage, hybrid dilated (HDRDAM) proposed second stage. The modules able learn appropriate local through connection between layers, irrelevant also be suppressed. sub-networks then connected by long skip retain shallow feature enhance experiments on seven benchmark datasets verified that compared with state-of-the-art methods, TSP-RDANet obtain favorable results both synthetic real denoising. code our available at https://github.com/WenCongWu/TSP-RDANet.