Deep Laplacian Pyramid Networks for Fast and Accurate Super-Resolution

By Wei-Sheng Lai, Jia-Bin Huang, Narendra Ahuja, Ming-Hsuan Yang. 2017

2,816 citations recorded by the source corpus.

Convolutional neural networks have recently demonstrated high-quality reconstruction for single-image super-resolution. In this paper, we propose the Laplacian Pyramid Super-Resolution Network (LapSRN) to progressively reconstruct the sub-band residuals of high-resolution images. At each pyramid level, our model takes coarse-resolution feature maps as input, predicts the high-frequency residuals, and uses transposed convolutions for upsampling to the finer level. Our method does not require the bicubic interpolation as the pre-processing step and thus dramatically reduces the computational complexity. We train the proposed LapSRN with deep supervision using a robust Charbonnier loss function and achieve high-quality reconstruction. Furthermore, our network generates multi-scale predictions in one feed-forward pass through the progressive reconstruction, thereby facilitates resource-aware applications. 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 speed and accuracy.

Subfields: Computer Vision and Pattern Recognition, Media Technology, Generative Models

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

In Generative Models

Citations in the corpus

Designing a Practical Degradation Model for Deep Blind Image Super-Resolution (2021, methodology): “As mentioned in the introduction, bicubic downsampling and traditional degradations underly existing DNNsbased SISR methods [22, 34, 41, 55], or some simple variants do [10, 11, 38, 47, 53, 54].”
Why Should We Add Early Exits to Neural Networks? (2020, methodology): “Apart from Inception, similar approaches have been successfully applied to, among others, face recognition [61], object detection [14], image segmentation [15], pose estimation [25], saliency detection [41], metric learning [71], visual attention prediction [69], and super resolution [34, 65].”
Hierarchical Neural Architecture Search for Single Image Super-Resolution (2020, background): “These two kinds of blocks/layers often follow a two-level architecture, where the network-level architecture determines the positions of the upsampling layers ( e.g. , SRCNN [7] and LapSRN [8]) and the cell-level architecture controls the computation of each block/layer ( e.g. , RCAB [9]).”
Deep Image Prior (2017, background): “bicubic upsampling, our method never has access to any data other than a single low-resolution image, and yet it produces much cleaner results with sharp edges close to state-of-the-art superresolution methods (LapSRN [18], SRResNet [19]) which utilize networks trained from large datasets.”
Image Super-Resolution Using Very Deep Residual Channel Attention Networks (2018, methodology): “LapSRN [6] MSLapSRN [7] ENet-PAT [8] MemNet [9] EDSR [10] SRMDNF [11] RCAN (ours)”
A Deep Journey into Super-resolution (2019, background): “Deep Laplacian Pyramid Super-Resolution Network (LapSRN) [47] employs a pyramidal framework.”
DRIT++: Diverse Image-to-Image Translation via Disentangled Representations (2019, background): “…can be formulated as I2I translation problems, such as colorization (Larsson et al. 2016; Zhang et al. 2016) (grayscale→ color), super-resolution (Lai et al. 2017; Ledig et al. 2017; Li et al. 2016, 2019) (low-resolution → high-resolution), and photorealistic image synthesis (Chen and Koltun…”
MambaIR: A Simple Baseline for Image Restoration with State-Space Model (2024, background): “Early attempts usually elaborate CNNs with techniques such as residual connection [6, 34], dense connection [68, 89] and others [13, 19, 36, 70] to improve model representation ability.”
BasicVSR: The Search for Essential Components in Video Super-Resolution and Beyond (2020, methodology): “We use Charbonnier loss [3] since it better handles outliers and improves the performance over the conventional `2 loss [19].”
Efficient Long-Range Attention Network for Image Super-resolution (2022, methodology): “The ELAN can be optimized with the commonly used loss functions for SR, such as L 2 [11, 20, 48, 49], L 1 [22, 29, 62] and perceptual losses [16, 45].”
End-to-end optimization of optics and image processing for achromatic extended depth of field and super-resolution imaging (2018, methodology): “We also apply a state-of-the art deep learning approach for single-image super-resolution [Lai et al. 2017] to the low-resolution image.”
Patch-Based Progressive 3D Point Set Upsampling (2018, methodology): “A recent plethora of deep learning super-resolution techniques have achieved significant improvement in single image super-resolution performance [9, 27, 32, 47]; in particular, multi-step methods have been shown to excel in their performance [11, 30, 62].”
Learning Enriched Features for Fast Image Restoration and Enhancement (2022, background): “…designed to perform SR include recursive learning [64], [65], [66], progressive reconstruction [67], [68], dense connections [7], [69], [70], attention mechanisms [71], [72], [73], multi-branch learning [68], [74], [75], [76], and generative adversarial networks (GANs) [70], [77], [78], [79].”
Training-Free, Single-Image Super-Resolution Using a Dynamic Convolutional Network (2018, neutral): “reconstructor network [22].”
Image denoising using deep CNN with batch renormalization (2020, background): “Speciﬁcally, the recently proposed deep CNNs have been widely applied in image restoration [33].”
Component Divide-and-Conquer for Real-World Image Super-Resolution (2020, positive): “, VDSR [10], EDSR [14], SRRestNet [13], LapSR [12] and RCAN [33], are presented to further improve the SR performance.”