To address the problems of insufficient low-rank constraints and high computational cost in traditional low⁃rank regularization Poisson denoising model, this paper proposed a Poisson denoising model based on Schatten⁃p norm low⁃rank regularization. Firstly, the noise image was segmented by adopting an overlapping blocking strategy, and clustering processing was conducted on the obtained image blocks. Meanwhile, the low‑rank properties of similar image blocks after clustering were leveraged, and the non‑convex Schatten⁃p norm was adopted as the regularization term to more accurately constrain the low⁃rank structure of similar block matrices. Secondly, matrix factorization was performed on the image blocks to reduce data storage, avoid singular value decomposition, and improve computational efficiency. Finally, due to the non‑convex and non‑smooth properties of the proposed model, the proximal alternating linearized minimization (PALM) algorithm was adopted for solution, which provided the convergent guarantee. Experimental results show that the proposed Poisson denoising model achieves an average improvement of 1.1 dB in the peak signal⁃to⁃noise ratio (PSNR) and 6.1% in the structural similarity index (SSIM), and its denoising performance is significantly superior to that of comparative algorithms.
DONATHA, SIEMIGINOWSKAA, KASHYAPV L, et al. Joint deconvolution of astronomical images in the presence of Poisson noise[J]. The Astronomical Journal, 2024, 168(4): 182.
[3]
ZHANGY D, ZHUY H, NICHOLSE, et al. A Poisson-Gaussian denoising dataset with real fluorescence microscopy images[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach: IEEE, 2019: 11702-11710.
[4]
LEEY, LEES, PARKC. Deep residual learning with Anscombe transformation for low-dose digital tomosynthesis[J]. Journal of the Korean Physical Society, 2024, 85(4): 333-341.
[5]
TANGJ S, FIENUPJ R. Improving accuracy of phase retrieval with mixed Poisson-Gaussian noise using the generalized Anscombe transform[C]//Imaging and Applied Optics Congress. Washington: Optica Publishing Group, 2020: CF2C-4.
[6]
SCHMIDTT, XUL. Some limit results on the haar-fisz transform for inhomogeneous Poisson signals[J]. Zeitschrift Für Analysis und Ihre Anwendungen, 2008, 27(4): 483-497.
[7]
SALMONJ, DELEDALLEC A, WILLETTR, et al. Poisson noise reduction with non-local PCA[C]//2012 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Kyoto: IEEE, 2012: 1109-1112.
[8]
SALMONJ, HARMANYZ, DELEDALLEC A, et al. Poisson noise reduction with non-local PCA[J]. Journal of Mathematical Imaging and Vision, 2014, 48(2): 279-294.
[9]
ZHAZ Y, WENB H, YUANX, et al. Nonlocal structured sparsity regularization modeling for hyperspectral image denoising[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 5510316.
[10]
ZHUX L, FENGX C, WANGW W, et al. HCLR: A hybrid clustering and low-rank regularization-based method for photon-limited image restoration[J]. Journal of Visual Communication and Image Representation, 2018, 57: 61-68.
[11]
XIEY, GUS H, LIUY, et al. Weighted Schatten $p$-norm minimization for image denoising and background subtraction[J]. IEEE Transactions on Image Processing, 2016, 25(10): 4842-4857.
[12]
ZHANGJ, LUJ J, WANGC, et al. Hyperspectral and multispectral image fusion via superpixel-based weighted nuclear norm minimization[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 5521612.
[13]
ZHANGM Y, ZHANGM L, ZHAOF, et al. Truncated weighted nuclear norm regularization and sparsity for image denoising[C]//2023 IEEE International Conference on Image Processing (ICIP). Kuala Lumpur: IEEE, 2023: 1825-1829.
[14]
GeR, LEEJ, MAT. Matrix completion has no spurious local minimum[J]. Advances in Neural Information Processing Systems, 2016, 29:2900-3627.
[15]
ZHENGJ W, QINM J, ZHOUX L, et al. Efficient implementation of truncated reweighting low-rank matrix approximation[J]. IEEE Transactions on Industrial Informatics, 2020, 16(1): 488-500.
[16]
HOUH J, NINGF L, LIW X, et al. Coherent noise denoising in beamforming based on non-convex robust principal component analysis[J]. Journal of Theoretical and Computational Acoustics, 2024, 32: 2340004.
[17]
CHENY Y, GUOY W, WANGY L, et al. Denoising of hyperspectral images using nonconvex low rank matrix approximation[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(9): 5366-5380.
[18]
JIAX X, FENGX C, WANGW W, et al. Generalized unitarily invariant gauge regularization for fast low-rank matrix recovery[J].IEEE Transactions on Neural Networks and Learning Systems, 2020, 99:1-15.
[19]
BOLTEJ, SABACHS, TEBOULLEM. Proximal alternating linearized minimization for nonconvex and nonsmooth problems[J]. Mathematical Programming, 2014, 146(1): 459-494.
[20]
HARMANYZ T, MARCIAR F, WILLETTR M. This is SPIRAL‑TAP: Sparse poisson intensity reconstruction ALgorithms-theory and practice[J]. IEEE Transactions on Image Processing, 2012, 21(3): 1084-1096.
[21]
JIAX X, FENGX C, WANGW W, et al. Online Schatten quasi-norm minimization for robust principal component analysis[J]. Information Sciences, 2019, 476: 83-94.
[22]
ERKANU, THANHD N H, HIEUL M, et al. An iterative mean filter for image denoising[J]. IEEE Access, 2019, 7: 167847-167859.