Fast and Accurate Image Super-Resolution with Deep Laplacian Pyramid Networks

By Wei-Sheng Lai, Jia-Bin Huang, Narendra Ahuja, Ming-Hsuan Yang. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017

908 citations recorded by the source corpus.

Convolutional neural networks have recently demonstrated high-quality reconstruction for single image super-resolution. However, existing methods often require a large number of network parameters and entail heavy computational loads at runtime for generating high-accuracy super-resolution results. In this paper, we propose the deep Laplacian Pyramid Super-Resolution Network for fast and accurate image super-resolution. The proposed network progressively reconstructs the sub-band residuals of high-resolution images at multiple pyramid levels. In contrast to existing methods that involve the bicubic interpolation for pre-processing (which results in large feature maps), the proposed method directly extracts features from the low-resolution input space and thereby entails low computational loads. We train the proposed network with deep supervision using the robust Charbonnier loss functions and achieve high-quality image reconstruction. Furthermore, we utilize the recursive layers to share parameters across as well as within pyramid levels, and thus drastically reduce the number of parameters. Extensive quantitative and qualitative evaluations on benchmark datasets show that the proposed algorithm performs favorably against the state-of-the-art methods in terms of run-time and image quality.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/0bbbe0ee7648a260275809699e146b6a2da57008

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

In Media Technology

Citations in the corpus

Multilevel and Multiscale Network for Single-Image Super-Resolution (2019, methodology): “To illustrate the effectiveness of our method, we compare the proposed method with other SR methods, including the Bicubic, SRCNN [10], VDSR [13], DRCN [26], MS-LapSRN [27], and SRMDNF [28] methods.”
Large Deformation Diffeomorphic Image Registration with Laplacian Pyramid Networks (2020, background): “The Laplacian pyramid network has demonstrated its efficiency and effectiveness in a variety of computer vision tasks, including high-resolution image synthetic [28,6], super-resolution [13] and optical flow estimation [29], in constructing highresolution solutions, stabilizing the training and avoi”
Convolutional Bipartite Attractor Networks (2019, background): “In a range of image-processing domains, e.g., denoising, inpainting, and super-resolution, performance gains are realized by constructing deeper and deeper architectures [e.g. , 20].”
Residual Dense Network for Image Restoration (2018, methodology): “MemNet [18] MSLapSRN [56] EDSR [23] D-DBPN [34] RDN (ours)”
Second-Order Attention Network for Single Image Super-Resolution (2019, methodology): “Some loss functions have been widely used, such as L 2 [2, 12, 29, 30], L 1 [14, 15, 20, 39], perceptual losses [11, 26].”
Transformer for Single Image Super-Resolution (2021, neutral): “FLOPs (x4) Running time VDSR [18] 20 Yes 0.67M 612.6G 0.00597s LapSRN [20] 27 Yes 0.25M 149.4G 0.00330s DRRN [33] 52 No 0.30M 6796.9G 0.08387s CARN [2] 34 Yes 1.6M 90.9G 0.00278s IMDN [16] 34 Yes 0.7M”
NAS-DIP: Learning Deep Image Prior with Neural Architecture Search (2020, background): “, single image super-resolution [44,45,46], image denoising [12,49,82], joint filtering [50]), and image synthesis (e.”
Efficient Transformer for Single Image Super-Resolution (2021, methodology): “1) Objective Evaluation: In TABLE I, we compare our ESRT with 13 state-of-the-art lightweight SR models, including SRCNN [14], FSRCNN [15], VDSR [16], DRCN [3], LapSRN [43], DRRN [2], MemNet [4], EDSR-baseline [17], SRMDNF [44], CARN [7], FALSR [5], IMDN [11], and LatticeNet [23].”
SNR-Aware Low-light Image Enhancement (2022, background): “, the Charbonnier loss [25] and the perceptual loss.”
Exploiting Aliasing for Manga Restoration (2021, methodology): “Many follow-ups reached a more accurate HR image by designing better network architectures, such as VDSR [9], SRResNet[12], LapSRN [11], or more powerful loss functions, like EDSR[14].”
A Deep Journey into Super-resolution (2019, methodology): “They also propose a single model, termed as Multi-Scale (MS) LapSRN, that jointly learns to handle multiple SR scales [48].”
Lightweight Image Super-Resolution with Information Multi-distillation Network (2019, methodology): “Some other work such as MS-LapSRN [15] and progressive SR (ProSR) [29] also adopt this progressive upsampling SR framework and achieve relatively high performance.”
Image Super-Resolution via Progressive Cascading Residual Network (2018, methodology): “Among them, the most successful one is LapSRN [17, 18].”
Deep Learning for Image Super-Resolution: A Survey (2019, methodology): “[77] feature similarity MS-SSIM [74] multi-scale SSIM IFC [75] information ﬁdelity criterion PSNR peak signal-to-noise ratio IQA image quality assessment SSIM [62] structural similarity MOS mean opinion score VIF [76] visual information ﬁdelity MSSIM [62] mean SSIM BN [140] batch normalization GAN [”
Densely Residual Laplacian Super-Resolution (2019, methodology): “Furthermore, this limitation is common to all the super-resolution methods [6], [7], [9], [11].”
Unsupervised Misaligned Infrared and Visible Image Fusion via Cross-Modality Image Generation and Registration (2022, methodology): “We compute L edge using the Charbonnier Loss [Lai et al. , 2019].”