Quick navigation

• repositories
• Datasets
• papers
  • Non-DL based approach
  • DL based approach
  • 2014-2016
  • 2017
  • 2018
  • 2019
  • 2020
  • 2021
• Super Resolution workshop papers
• Super Resolution survey

Awesome-Super-Resolution（in progress）

Collect some super-resolution related papers, data and repositories.

papers

DL based approach

Note this table is referenced from here

2021

More years papers, plase check Quick navigation

| Title | Model | Published | Code | Keywords | | ---------------------- | ---------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | |Trilevel Neural Architecture Search for Efficient Single Image Super-Resolution | TrilevelNAS | arxiv | - | Trilevel Architecture Search Space | |SplitSR: An End-to-End Approach to Super-Resolution on Mobile Devices | SplitSR | arxiv | - | lightweight,on Mobile Devices | |Learning for Unconstrained Space-Time Video Super-Resolution | USTVSRNet | arxiv | - | VSR, Unconstrained video super-resolution,general-ized pixelshuffle layer. | |ClassSR: A General Framework to Accelerate Super-Resolution Networks by Data Characteristic| ClassSR | cvpr21 | code | classification，lightweight | |Collapsible Linear Blocks for Super-Efficient Super Resolution | SESR | arxiv | - | Super-Efficient SR, overparameterization| |Self-Supervised Adaptation for Video Super-Resolution | Adapted VSR | arxiv | - | VSR, Self-Supervised Adaptationn| |Generic Perceptual Loss for Modeling Structured Output Dependencies | Generic Perceptual Loss| cvpr21 | - | Random VGG w/o pretrianed, work at semantic segmentation, depthestimation and instance segmentation| |Large Motion Video Super-Resolution with Dual Subnet and Multi-Stage Communicated Upsampling| U3D-RDN | AAAI21 | - | VSR, Dual Subnet, Multi-stage Communicated Up-sampling| |UltraSR: Spatial Encoding is a Missing Key for Implicit Image Function-based Arbitrary-Scale Super-Resolution| UltraSR | arxiv | - | Implicit Image Function Based, Arbitrary-Scale| |D2C-SR: A Divergence to Convergence Approach for Image Super-Resolution | D2C-SR | arxiv | - | RealSR， divergence stage with a triple loss ，convergence stage| |Training a Better Loss Function for Image Restoration | MDF loss | arxiv | code | Multi-Scale Discriminative Feature loss| |Transitive Learning: Exploring the Transitivity of Degradations for Blind Super-Resolution| TLSR | arxiv | code | Blind SR， Transitive Learning | |Best-Buddy GANs for Highly Detailed Image Super-Resolution | Beby-GAN | arxiv | code |relaxing the immutable one-to-one constraint, Best-Buddy Loss | |Flow-based Kernel Prior with Application to Blind Super-Resolution | FKP | cvpr21 | code |Blind SR, flow-based kernel prio| |Conditional Meta-Network for Blind Super-Resolution with Multiple Degradations | CMDSR | arxiv | - |Blind SR, Conditional Meta-Network| |Image Super-Resolution via Iterative Refinement | SR3 | arxiv | - |Repeated Refinement, better than SOTA GAN| |BAM: A Lightweight and Efficient Balanced Attention Mechanism for Single Image Super Resolution| BAM | arxiv | code |balanced Attention Mechanism | |Kernel Agnostic Real-world Image Super-resolution | KASR | arxiv | - |realsr, Kernel Agnostic | |Neural Architecture Search for Image Super-Resolution Using Densely Constructed Search Space: DeCoNAS| DeCoNAS | arxiv | - |Densely Constructed Search Space | |Attention in Attention Network for Image Super-Resolution | A2N | arxiv | code |Attention in Attention， lightweight | |Temporal Modulation Network for Controllable Space-Time Video Super-Resolution | TMNet | arxiv | code | VSR， interpolate frames, Temporal Modulation Block | |A Two-Stage Attentive Network for Single Image Super-Resolution | TSAN | arxiv | code | SISR， multi-context attentiveblock | |BasicVSR++: Improving Video Super-Resolution with Enhanced Propagation and Alignment | BasicVSR++ | arxiv | code | VSR， 3 champions and 1 runner-up in NTIRE 2021 | |SRDiff: Single Image Super-Resolution with Diffusion Probabilistic Models | SRDiff | arxiv | | Diffusion Probabilistic Models | |SRWarp: Generalized Image Super-Resolution under Arbitrary Transformation | SRWarp | cvpr21 |code | arbitrary transformation | |Cross-MPI: Cross-scale Stereo for Image Super-Resolutionusing Multiplane Images | Cross-MPI | cvpr21 |- | RefSR, plane-aware attention, coarse-to-fine guided upsampling | |Lightweight Image Super-Resolution with Hierarchical and DifferentiableNeural Architecture Search | DLSR | arxiv |code | Lightweight | |HINet: Half Instance Normalization Network for Image Restoration | HINet | arxiv |- | Half Instance Normalization Block, 1st place on the NTIRE 2021 Image Deblurring Challenge - Track2. JPEG Artifacts | |FDAN: Flow-guided Deformable Alignment Network for Video Super-Resolution | FDAN | arxiv |- | VSR, Flowguided Deformable Module | |End-to-end Alternating Optimization for Blind Super Resolution | DAN | arxiv |code | Blind SR, Restorer and Estimator Alternating Optimization | |Anchor-based Plain Net for Mobile Image Super-Resolution | ABPN | arxiv |code | MAI2021, mobile device SISR, INT8 Quantization | |Extremely Lightweight Quantization RobustReal-Time Single-Image Super Resolution for Mobile Devices | XLSR | arxiv |code | MAI2021, mobile device SISR Winner, INT8 Quantization | |Robust Reference-based Super-Resolution via C2-Matching | C2-Matching | arxiv |code | RefSR, ransformation gap, contrastive correspondence network, resolution gap, teacher-student correlation distillation | |MASA-SR: Matching Acceleration and Spatial Adaptation forReference-Based Image Super-Resolution | MASA-SR | cvpr21 |code | RefSR, Match & Extraction Module , Spatial Adaptation Module| |Noise Conditional Flow Model for Learning the Super-Resolution Space | NCSR | arxiv |code | better than GAN-based model，Flow-based， noise conditional layer | |Variational AutoEncoder for Reference based Image Super-Resolution | RefVAE | arxiv |code | Variational AutoEncoder， refsr | |Video Super-Resolution Transformer | VSR-Transformer | arxiv |code | VSR, spatial-temporal convolutional self-attention layer, bidirectional optical flow-based feed-forward layer | |Practical Single-Image Super-Resolution Using Look-Up Table | SR-LUT | cvpr21 |- | SISR, fater than Bicubic, look-up table | |Towards Fast and Accurate Real-World Depth Super-Resolution: BenchmarkDataset and Baseline | FDSR | cvpr21 |- | depth map SR | |Image Super-Resolution with Non-Local Sparse Attention | NLSN | cvpr21 |- | Non-Local Sparse Attention， Spherical LSH | |Learning the Non-differentiable Optimization for Blind Super-Resolution | AMNet, AMGAN | cvpr21 |- | Blind SR， non-differentiable, adap-tive modulation network | |KOALAnet: Blind Super-Resolution using Kernel-Oriented Adaptive Local Adjustment | KOALAnet | cvpr21 |- | Blind SR, kernel-oriented adaptive local adjustment, learns spatially-variant degradation and restoration kernels | |Exploring Sparsity in Image Super-Resolution for Efficient Inference| SMSR | cvpr21 |code| Sparse Masks, Efficient SISR| |LAU-Net: Latitude Adaptive Upscaling Network for Omnidirectional Image Super-resolution| LAU-Net | cvpr21 |code| Omnidirectional Image SR| |Tackling the Ill-Posedness of Super-Resolution through Adaptive Target Generation| AdaTarget | cvpr21 |code| Adaptive Target Generator| |Single Pair Cross-Modality Super Resolution| CMSR | cvpr21 | - | single-pair,Cross-Modality | |End-to-End Learning for Joint Image Demosaicing, Denoising and Super-Resolution| JDNDMSR | cvpr21 | - | joint tasks | |MR Image Super-Resolution with Squeeze and Excitation Reasoning Attention| SERAN | cvpr21 | - | MRI SR, squeeze and excitation reasoning attention networks | |Light Field Super-Resolution with Zero-Shot Learning| - | cvpr21 | - | zero shot, light field SR | |Scene Text Telescope: Text-Focused Scene Image Super-Resolution| TBSRN | cvpr21 | - | text-focused SR | |Interpreting Super-Resolution Networks with Local Attribution Maps| LAM | cvpr21 | - | Interpreting SR,local attribution map | |Turning Frequency to Resolution: Video Super-resolution via Event Cameras| E-VSR | cvpr21 | - | VSR, Event-based VSR | |GLEAN: Generative Latent Bank for Large-Factor Image Super-Resolution |GLEAN | cvpr21 | - | Latent Bank, large scale factor | |BasicVSR: The Search for Essential Components in Video Super-Resolution and Beyond |BasicVSR | cvpr21 | - | VideoSR, The Search for Essential Components | |AdderSR: Towards Energy Efficient Image Super-Resolution|AdderSR | cvpr21 |- | SISR, adder neural networks, Energy Efficient | |Deep Burst Super-Resolution | BurstSR | cvpr21 | - | multi-frame sr, new BurstSR dataset | |Pre-Trained Image Processing Transformer |IPT | cvpr21 |code | Pre-Trained Image Processing Transformer, Imagenet pretrained, dramatically improve performance | |Blind Image Super-Resolution via ContrastiveRepresentation Learning | CRL-SR | arxiv |- | blind SR, contrastive decoupling encoding， contrastive feature refinement | |End-to-End Adaptive Monte Carlo Denoising and Super-Resolution | - | arxiv |- | Monte Carlo path tracing | |SwinIR: Image Restoration Using Swin Transformer | SwinIR | arxiv |code |SISR， Swin Transformer， SOTA | |edge–SR: Super–Resolution For The Masses | eSR | arxiv |- |SISR， Edge device SR，one–layer architectures use interpretable mechanisms， Filling the gap between classic and deep learning architectures | |Memory-Augmented Non-Local Attention for Video Super-Resolution | CSNLN | arxiv |- |VSR， without frame alignment， memory-augmented attention module | |Simple and Efficient Unpaired Real-world Super-Resolution using Image Statistics | - | arxiv |- |unpair SR,variance matching | |Improving Super-Resolution Performance using Meta-Attention Layers | - | SPL |code | SISR， meta-attention | |EFENet: Reference-based Video Super-Resolution with Enhanced Flow Estimation | EFENet | arxiv |code | RefVSR， Enhanced Flow Estimation | |Learning A Single Network for Scale-Arbitrary Super-Resolution | arbSR | iccv21 |code | SISR， scale-arbitrary SR | |Dual-Camera Super-Resolution with Aligned Attention Modules | - | iccv21 |code | oral, dual camera SR, RefSR,self-supervised domain adaptation strategy | |Learning Frequency-aware Dynamic Network for Efficient Super-Resolution | FADN | iccv21 |- | Efficient SR，DCT，Mask Predictor，dynamic resblocks，frequency mask loss | |Designing a Practical Degradation Model for Deep Blind Image Super-Resolution | BSRNet/BSRGAN | iccv21 |- | randomly shuffled blur, downsampling and noise degradations for degradation model | |Fourier Space Losses for Efficient Perceptual Image Super-Resolution | - | iccv21 |- | Fourier space supervision loss | |Omniscient Video Super-Resolution | OVSR | iccv21 |- | VSR， new framework， precursor net and successor net | |Efficient Video Compression via Content-Adaptive Super-Resolution | SRVC | iccv21 |- | Video Compression， Adaptive Conv | |Mutual Affine Network for Spatially Variant Kernel Estimation in Blind Image Super-Resolution | MANet | iccv21 |code | blind SR, spatially variant and invariant kernel, mutual affine network | |Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling | HCFlow | iccv21 |code | SR/Rescale, learns a bijective mapping | |Deep Reparametrization of Multi-Frame Super-Resolution and Denoising | - | iccv21 |- | Multi-Frame，deep reparametrization of the classical MAP objective | |Attention-based Multi-Reference Learning for Image Super-Resolution | AMRSR | iccv21 |code | RefSR， without frame alignment， Hierarchical Attention-based Similarity | | COMISR: Compression-Informed Video Super-Resolution | COMISR | iccv21 | code | VSR, bi-directional recurrent, warping, detail-preserving flow estimation, Laplacian enhancement | |Achieving on-Mobile Real-Time Super-Resolution with Neural Architecture and Pruning Search | - | iccv21 |- | SISR, real-time sr, NAS | |Event Stream Super-Resolution via Spatiotemporal Constraint Learning | - | iccv21 |- | event stream SR | |Dynamic High-Pass Filtering and Multi-Spectral Attention for Image Super-Resolution | DFSA | iccv21 |- | SISR, SOTA, matrix multi-spectral channel attention | |Context Reasoning Attention Network for Image Super-Resolution | CRAN | iccv21 |- | SISR, SOTA, context reasoning attention | |EvIntSR-Net: Event Guided Multiple Latent Frames Reconstruction and Super-Resolution | - | iccv21 |- | EvIntSR | |Deep Blind Video Super-Resolution | - | iccv21 |- | Blind VSR | |Super Resolve Dynamic Scene From Continuous Spike Streams | MGSR | iccv21 |- | Spike camera SR | |Benchmarking Ultra-High-Definition Image Super-Resolution | MANet | iccv21 |- | UHD SR dataset, 4K, 8K, SISR | |Lucas-Kanade Reloaded: End-to-End Super-Resolution from Raw Image Bursts | - | iccv21 |code | Raw Image Bursts SR | |Unsupervised Real-World Super-Resolution: A Domain Adaptation Perspective | - | iccv21 |- | unpaired realSR, domain adaptation | |Real-world Video Super-resolution: A Benchmark Dataset and A Decomposition based Learning Scheme | - | iccv21 |code | RealVSR dataset | |Scale-Aware Dynamic Network for Continuous-Scale Super-Resolution | SADN | arxiv |code | Arbitrary scale SR, scale-aware dynamic conv | |MEGAN: Memory Enhanced Graph Attention Network for Space-Time Video Super-Resolutio | MEGAN | arxiv | - | space-time VSR, long-range memory graph aggregation | |Normalizing Flow as a Flexible Fidelity Objective for Photo-Realistic Super-resolution | - | arxiv | code | flow-based NLL loss,replace L1 | |Frequency-Aware Physics-Inspired Degradation Model for Real-World Image Super-Resolution | - | arxiv | - | learn the cutoff frequency of real-world degradation process | |Local-Selective Feature Distillation for Single Image Super-Resolution | - | arxiv | - | feature distillation , local-selective feature distillation (LSFD) |Investigating Tradeoffs in Real-World Video Super-Resolution | RealBasicVSR | arxiv | code | RealVSR | |A Practical Contrastive Learning Framework for Single Image Super-Resolution | - | arxiv | - | SISR, Task-Generalizable Embedding | |AdaDM: Enabling Normalization for Image Super-Resolution | AdaDM | arxiv | code | SISR, apply BN in SR model | |Revisiting Temporal Alignment for Video Restoration | - | arxiv | code |VSR, iterative alignment module |