Video Denoising, Deblocking, and Enhancement Through Separable 4-D Nonlocal Spatiotemporal Transforms

By Matteo Maggioni, Giacomo Boracchi, Alessandro Foi, Karen O. Egiazarian. IEEE Transactions on Image Processing, 2012

408 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/ead8f6a86022d998fc4e788985f24b892b7d4552

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Burst Denoising with Kernel Prediction Networks (2017, methodology): “(d) Non-local means [3] (e) VBM4D [17] (f) Our KPN model”
TDAN: Temporally-Deformable Alignment Network for Video Super-Resolution (2018, methodology): “Generic: The proposed TDAN is a generic temporal alignment framework and can be easily utilized to replace ﬂow-based motion compensation for several other tasks, such as video denoising [19], video deblocking [18], video deblur-ring [8], video frame interpolation [1], and even video prediction [30].”
Video Enhancement with Task-Oriented Flow (2017, methodology): “.8759 Fixed ﬂow 28.38 0.9333 28.56 0.9200 — — — — TOFlow 29.10 0.9544 28.85 0.9407 29.41 0.9271 30.36 0.8855 Table 2: Quantitative comparisons on video denoising Methods Vimeo-Blocky PSNR SSIM V-BM4D [20] 35.75 0.9587 Fixed ﬂow 36.52 0.9636 TOFlow 36.92 0.9663 Table 3: Results on video deblocking In”
FastDVDnet: Towards Real-Time Deep Video Denoising Without Flow Estimation (2019, methodology): “In this context, motion estimation and/or compensation has been employed in a number of video denoising algorithms to help to improve denoising performance and temporal consistency [22, 38, 2, 25, 4].”
Minimum Noise Fraction versus Principal Component Analysis as a Preprocessing Step for Hyperspectral Imagery Denoising (2016, methodology): “It can be seen that the PCA+VBM4D achieves the highest denoising results among all methods for this noise scenario.”
MEMC-Net: Motion Estimation and Motion Compensation Driven Neural Network for Video Interpolation and Enhancement (2018, methodology): “19, the blocky regions around the hand and eye are sufficiently reduced by both MEMCNet_DB and V-BM4D [58] methods.”
VRT: A Video Restoration Transformer (2022, background): “For example, [44] and [45] exploited the non-local means in video denoising, by searching the pixels that resemble the pixel in restoration.”
Seeing Motion in the Dark (2019, background): “VBM4D [32] can remove the noise to an extent, but cannot correct the color shift.”
Fast Hyperspectral Image Denoising and Inpainting Based on Low-Rank and Sparse Representations (2018, background): “Identical ideas have been pursued in multiband image denoising: BM4D [12], VBM4D [13], and MSPCA-BM3D [14] use collaborative filtering in groups of three-dimensional (3-D) patches extracted from volumetric data, videos, multispectral data, respectively.”
Rank Minimization for Snapshot Compressive Imaging (2018, methodology): “More speciﬁcally, DeSCI-WNNM outperforms DeSCI-VBM4D more than 2 dB on PSNR and more than 0.025 on SSIM.”