Remote Sensing Image Denoising via Low-Rank Tensor Approximation and Robust Noise Modeling

Abstract

Noise removal is a fundamental problem in remote sensing image processing. Most existing methods, however, have not yet attained sufficient robustness in practice, due to more or less neglecting the intrinsic structures of remote sensing images and/or underestimating the complexity of realistic noise. In this paper, we propose a new remote sensing image denoising method by integrating intrinsic image characterization and robust noise modeling. Specifically, we use low-Tucker-rank tensor approximation to capture the global multi-factor correlation within the underlying image, and adopt a non-identical and non-independent distributed mixture of Gaussians (non-i.i.d. MoG) assumption to encode the statistical configurations of the embedded noise. Then, we incorporate the proposed image and noise priors into a full Bayesian generative model and design an efficient variational Bayesian algorithm to infer all involved variables by closed-form equations. Moreover, adaptive strategies for the selection of hyperparameters are further developed to make our algorithm free from burdensome hyperparameter-tuning. Extensive experiments on both simulated and real multispectral/hyperspectral images demonstrate the superiority of the proposed method over the compared state-of-the-art ones.

Texture-Sensitive Superpixeling and Adaptive Thresholding for Effective Segmentation of Sea Ice Floes in High-Resolution Optical Images

“…In fact, as remote sensing images are characterized with low-rankness, i.e. sparsity, which means that the high-dimensional data can be adequately represented in a low-dimensional subspace for dimensionality reduction [31]. Basically, the noise component tends to be sparse and lies in the sparse subspace on the dataset.…”

Hyperspectral Image Restoration Combining Intrinsic Image Characterization With Robust Noise Modeling

“…Prior of the noise E. Inspired by our previous work [34], we impose the following non-i.i.d. MoG prior on E, which assumes that noise in different bands follows distinct and correlated…”

“…noise priors such as non-i.i.d. MoG [12,34] and Dirichlet process Gaussian mixture [56], showing robust noise fitting capability in realistic scenarios. [18] Matrix LR Spatial TV None (inpainting) HaLRTC [31] Tensor Tucker LR -None (inpainting) t-SVD [60] Tensor tubal LR -None (inpainting) KBR-TC [54] Tensor KBR LR -None (inpainting) WLRTR [10] Tensor HOSVD LR Self-similarity None (inpainting) NL-LRTC [24] Tensor Tucker LR Self-similarity None (inpainting) TVTR [21] Tensor ring LR Spatial TV None (inpainting) LRMR [58] Matrix LR -Gaussian+Laplacian NMoG-LRMF [12] Matrix LR -Non-i.i.d.…”

Guaranteed Robust Tensor Completion via ∗L-SVD with Applications to Remote Sensing Data

“…When only a fraction of partially corrupted observations are available, the crucial point of robust tensor completion lies in the assumption that the underlying data tensor is highly redundant such that the main components of it remain only slightly suppressed by missing information, outliers, and noises, and thus can be effectively reconstructed by exploiting the intrinsic redundancy. The tensor low-rankness is an ideal tool to model the redundancy of tensor data, and has gained extensive attention in remote sensing data restoration [5,10,11].…”