Image Denoising Via Sparse and Redundant Representations Over Learned Dictionaries
By Michael Elad, Michal Aharon. IEEE Transactions on Image Processing, 2006
5,637 citations recorded by the source corpus.
We address the image denoising problem, where zero-mean white and homogeneous Gaussian additive noise is to be removed from a given image. The approach taken is based on sparse and redundant representations over trained dictionaries. Using the K-SVD algorithm, we obtain a dictionary that describes the image content effectively. Two training options are considered: using the corrupted image itself, or training on a corpus of high-quality image database. Since the K-SVD is limited in handling small image patches, we extend its deployment to arbitrary image sizes by defining a global image prior that forces sparsity over patches in every location in the image. We show how such Bayesian treatment leads to a simple and effective denoising algorithm. This leads to a state-of-the-art denoising performance, equivalent and sometimes surpassing recently published leading alternative denoising methods
Subfields: Computational Mechanics, Computer Vision and Pattern Recognition, Media Technology
View Paper: https://www.semanticscholar.org/paper/e07416eabd4ba6c69fa473756bb04ae7161177be
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computational Mechanics
- Lei Zhang
- Julien Mairal
- Francis R. Bach
- Wangmeng Zuo
- Xiangchu Feng
- David Dapeng Zhang
- Jean Ponce
- Michael Elad
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Ming-Hsuan Yang
- Dacheng Tao
- Lei Zhang
- Xiaogang Wang
- Trevor Darrell
- Andrew Zisserman
- Jitendra Malik
In Media Technology
- Lei Zhang
- Wangmeng Zuo
- Radu Timofte
- Kai Zhang
- Ming-Hsuan Yang
- Luc Van Gool
- Chen Change Loy
- Shutao Li
Citations in the corpus
- Super-Resolution with Deep Convolutional Sufficient Statistics (2015, background): “Parametric models with sparse regularization have enjoyed great success in this problem (Elad & Aharon (2006); Yang et al. (2008)), thanks to their capacity to capture local regularity with efficient convex optimization methods (Mairal et al. (2009)).”
- MR Image Reconstruction From Highly Undersampled k-Space Data by Dictionary Learning (2011, methodology): “Recent studies on adaptive dictionaries [24], [25] have shown the promise of patch-based sparsifying dictionaries in a variety of applications such as image/video denoising, image/video inpainting, deblurring, and demosaicing [ 26 ]‐[31].”
- A New Pan-Sharpening Method Using a Compressed Sensing Technique (2011, background): “In recent years, the sparsity-regularization-based image processing methods, such as denoising [31], [32], face recognition [33], and superresolution [34], always lead to the state-of-the-art results.”
- The Cosparse Analysis Model and Algorithms (2011, neutral): “…(e.g. [25, 16, 41, 42, 2]), addressing ways to obtain the dictionary from a bulk of data (e.g. [22, 1, 30, 38]), and beyond all these, attacking a long series of applications in signal and image processing with this model, demonstrating often state-of-the-art results (e.g. [20, 18, 28, 34]).”
- Non-local Regularization of Inverse Problems (2008, background): “This leads to redundant representations that leads to state of the art image denoising results [22].”
- Group-Sparse Representation With Dictionary Learning for Medical Image Denoising and Fusion (2012, positive): “Following the previous work [11], the local patches are considered instead of the whole image.”
- Image denoising: Can plain neural networks compete with BM3D? (2012, methodology): “A conceptually similar approach is to denoise image patches by trying to approximate noisy patches using a sparse linear combination of elements of a learned dictionary [1, 4].”
- Solving Linear Inverse Problems Using Gan Priors: An Algorithm with Provable Guarantees (2018, methodology): “In [6, 7], sparse and redundant dictionaries are learned for image denoising, whereas in [25, 26, 8], total variation is used as a regularizer.”
- Pixel-level image fusion with simultaneous orthogonal matching pursuit (2012, background): “Thus, common image features can be well described by only ‘a few descriptions’ of the sparse representation coefficients [28].”
- Taking Human out of Learning Applications: A Survey on Automated Machine Learning (2018, background): “Popular examples are sparse coding [68] (and its convolutional variants [69]) and local-linear coding [70].”
- Sparse Representation for Color Image Restoration (2008, methodology): “The alternative path, a nonparametric learning, uses image examples directly within the reconstruction process, as practiced by Efros and Leung for texture synthesis [33], by Freeman et al. for super-resolution [34], [35], and by several follow-up works [36]–[39] for super-resolution and inpainting.”
- Generalized Domain-Adaptive Dictionaries (2013, background): “Since then, data-driven dictionaries have been shown to work well for both image restoration [3] and classification tasks [26].”
- Sparse Representation for Computer Vision and Pattern Recognition (2010, background): “, [17], [20], [30], have been developed in recent years, leading to state-ofthe-art results in many signal and image processing tasks [11], [32], [44], [48], [51], [54].”
- Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising (2016, background): “Over the past few decades, various models have been exploited for modeling image priors, including nonlocal self-similarity (NSS) models [1], [2], [3], [4], sparse models [4], [5], [6], gradient models [7], [8], [9] and Markov random field (MRF) models [10], [11], [12].”
- Learning a discriminative dictionary for sparse coding via label consistent K-SVD (2011, methodology): “Sparse coding has been successfully applied to a variety of problems in computer vision and image analysis, including image denoising [ 5 ], image restoration [22, 25], and image classification [29, 28, 4]. Sparse coding approximates an input signal, y, by a sparse linear combination of items from a”
- Plug-and-Play Image Restoration With Deep Denoiser Prior (2020, positive): “The second step is to solve the prior subproblem with any off-the-shelf denois-ers, such as K-SVD [21], non-local means [22], BM3D [23].”
- Seismic data restoration via data-driven tight frame (2014, methodology): “Under certain assumptions (Cai et al., 2013), one can implement this step much faster than other existing dictionary-learning algorithms (e.g., K-SVD [Elad and Aharon, 2006; Aharon et al., 2006]).”
- Solving Inverse Problems With Piecewise Linear Estimators: From Gaussian Mixture Models to Structured Sparsity (2010, background): “Recently, very impressive image restoration results have been obtained with local patch-based sparse representatio ns calculated with dictionaries learned from natural images [ 1], [24], [48], [51], [76].”