Coupled Dictionary Training for Image Super-Resolution

By Jianchao Yang, Zhaowen Wang, Zhe L. Lin, Scott D. Cohen, Thomas S. Huang. IEEE Transactions on Image Processing, 2012

825 citations recorded by the source corpus.

Subfields: Computational Mechanics, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/7aa18f27d6dd0237b7c2590d4be64170c984ec28

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

In Computer Vision and Pattern Recognition

Citations in the corpus

End-to-End Image Super-Resolution via Deep and Shallow Convolutional Networks (2016, methodology): “, sparse coding [4], [5], [7], anchored neighbor [8]–[11], regression trees or forests [12]–[14], have been adopted to learn the mapping function.”
Single image super-resolution from transformed self-exemplars (2015, methodology): “These include nearest neighbor [14], kernel ridge regression [23], sparse coding [41] [40] [42] [36], manifold learning [6] and convolutional neural networks [9].”
Generalized Domain-Adaptive Dictionaries (2013, background): “In the dictionary learning literature, Yang et al. [27] and Wang et al. [24] proposed learning dictionary pairs for cross-modal synthesis.”
Sparse Modeling for Image and Vision Processing (2014, methodology): “A strategy for overcoming the discrepancy between training and testing in the previous approach has been independently proposed by several authors [Zeyde et al., 2012, Couzinie-Devy et al., 2011, Yang et al., 2012a].”
Enhanced Deep Residual Networks for Single Image Super-Resolution (2017, methodology): “Those learning methods rely on techniques ranging from neighbor embedding [3, 2, 7, 21] to sparse coding [31, 32, 27, 33].”
Learning Transferable Knowledge Through Embedding Spaces (2019, methodology): “In this chapter, we focus on this latter issue of learning the mapping for ZSL, using the novel approach of coupled dictionary learning (CDL) [57] to relate the visual features and the semantic attributes.”
Image Super-Resolution Using Deep Convolutional Networks (2014, methodology): “These methods either exploit internal similarities of the same image [5], [13], [16], [19], [47], or learn mapping functions from external low- and high-resolution exemplar pairs [2], [4], [6], [15], [23], [25], [37], [41], [42], [47], [48], [50], [51].”
SRFeat: Single Image Super-Resolution with Feature Discrimination (2018, background): “neighborhood embedding [7, 45, 46, 36], sparse coding [52, 54, 51, 16], and local mapping function regression [15, 50, 38].”
Frame-Recurrent Video Super-Resolution (2018, methodology): “Super-resolution is a classic ill-posed inverse problem with approaches ranging from simple interpolation methods such as Bilinear, Bicubic and Lanczos [8] to example-based super-resolution [11, 12, 37, 39], dictionary learning [30, 40], and self-similarity approaches [15, 38].”
Super-Resolution with Deep Convolutional Sufficient Statistics (2015, background): “Recent works have shown that better results can be obtained when the models are tuned on recovery performance rather than on data ﬁtting (Yang et al. (2012); Mairal et al. (2012)).”
Image Super-Resolution via Dual-State Recurrent Networks (2018, background): “It demonstrated superior performance to many other example-based learning paradigms, such as nearest neighbor [13], sparse representation [44, 43], neighborhood embedding [5, 37], random forest [32], etc.”
Deeply Improved Sparse Coding for Image Super-Resolution (2015, methodology): “We do not list other methods [31, 25, 33, 15, 13] whose performance has been shown worse than A+ or CNN-L.”
EnhanceNet: Single Image Super-Resolution Through Automated Texture Synthesis (2016, neutral): “They further include dictionary-based methods [33, 40, 52, 59, 61, 64] that learn a sparse representation of image patches as a combination of dictionary atoms, as well as neural network-based approaches [4, 8, 9, 24, 25, 26, 47, 48, 62] which apply convolutional neural networks (CNNs) to the task…”
Real-Time Video Super-Resolution with Spatio-Temporal Networks and Motion Compensation (2016, methodology): “SR aims at recovering a high-resolution (HR) image or video from its low-resolution (LR) version, and ﬁnds direct applications ranging from medical imaging [38, 34] to satellite imaging [5], as well as facilitating tasks such as face recognition [13].”
Image Super-Resolution Using Dense Skip Connections (2017, background): “The dictionary can be formed by a database of patches or learnt from the database [27].”
Learning a Deep Convolutional Network for Image Super-Resolution (2014, methodology): “A category of state-of-the-art SR approaches [9,4,25,26,24,2,28,20] learn a mapping between low/high-resolution patches.”
Deep Networks for Image Super-Resolution with Sparse Prior (2015, methodology): “SR results given by SC [35] (first row), CNN [8] (second row) and our CSCN (third row).”