复旦大学数字信号处理与传输实验室的冯辉老师的合作论文Pixel-Patch Graph Regularized Group Sparse Representation for Single-Image Denoising，被2026 IEEE International Conference on Acoustics, Speech and Signal Processing (2026 ICASSP)录用

论文摘要：Group sparse representation (GSR) is a model-based unsupervised paradigm that has proved effective for single-image denoising. Most GSR-based methods focus on preserving similarity but often ignore noise with spatial correlations, resulting in over-smoothing. In this paper, pixel-patch graph regularized group sparse representation (PPGR-GSR) is proposed to address this limitation. We first introduce the dual-graph structure at the group level. Specifically, a patch graph models the relationships between patches to preserve structural patterns, and a pixel graph captures the information of multiple pixels at the same position in matched patches, preventing noise variations from diffusing disorderly. Then, Laplacian regularization is applied to the dual graph, ensuring that the sparse representations vary smoothly along the underlying data manifold. The problem is solved via the alternating direction method of multipliers (ADMM) strategy. Experimental results on synthetic and real datasets demonstrate that the proposed PPGR-GSR outperforms existing single-image denoising methods.

论文作者：Xinxin Hou, Xue-Qin Jiang, Shubo Zhou, Hui Feng