ICCV 2023 超分辨率（super-resolution）方向上接收论文总结

ICCV 2023

官网链接：https://iccv2023.thecvf.com/
会议时间：2023 年 10 月 2 日至 6 日，法国巴黎（Paris）。
ICCV 2023统计数据：收录 2160 篇。

现将超分辨率方向上接收的论文汇总如下，遗漏之处还请大家斧正。

图像超分

1. SRFormer: Permuted Self-Attention for Single Image Super-Resolution
   • Paper: http://arxiv.org/abs/2303.09735
   • Keywords: Transformer
   • Blog: ICCV 2023 | 南开程明明团队提出新颖注意力机制用于图像超分辨率任务
2. Dual Aggregation Transformer for Image Super-Resolution
   • Paper: https://arxiv.org/abs/2308.03364
   • Code: https://github.com/zhengchen1999/DAT
   • Keywords: Transformer
   • Video: ICCV 2023【已开源】| 上交大与苏黎世联邦理工学院领衔发布，即插即用双向聚合Transformer
   • Blog: ICCV 2023 | 利用双重聚合的Transformer进行图像超分辨率
3. Feature Modulation Transformer: Cross-Refinement of Global Representation via High-Frequency Prior for Image Super-Resolution
   • Paper: http://arxiv.org/abs/2308.05022
   • Code: https://github.com/AVC2-UESTC/CRAFT-SR.git
   • Keywords: Transformer
4. MSRA-SR: Image Super-resolution Transformer with Multi-scale Shared Representation Acquisition
   • Paper: ICCV 2023 Open Access Repository
   • Code:
   • Keywords: Transformer
5. Content-Aware Local GAN for Photo-Realistic Super-Resolution
   • Paper: ICCV 2023 Open Access Repository
   • Code: https://github.com/jkpark0825/CAL
   • Keywords: GAN

轻量化超分

1. Lightweight Image Super-Resolution with Superpixel Token Interaction
   • Paper: ICCV 2023 Open Access Repository
   • Code: https://github.com/ArcticHare105/SPIN
   • Keywords: Lightweight
2. Iterative Soft Shrinkage Learning for Efficient Image Super-Resolution
   • Paper: ICCV 2023 Open Access Repository
   • Code: https://github.com/Jiamian-Wang/Iterative-Soft-Shrinkage-SR
   • Keywords: Efficient
3. Reconstructed Convolution Module Based Look-Up Tables for Efficient Image Super-Resolution
   • Paper: https://arxiv.org/abs/2307.08544
   • Code: https://github.com/liuguandu/RC-LUT
   • Keywords: Efficient
4. Boosting Single Image Super-Resolution via Partial Channel Shifting
   • Paper: ICCV 2023 Open Access Repository
   • Code: https://github.com/OwXiaoM/PCS
   • Keywords: Efficient
5. Spatially-Adaptive Feature Modulation for Efficient Image Super-Resolution
   • Paper: https://arxiv.org/abs/2302.13800
   • Code: https://github.com/sunny2109/SAFMN
   • Keywords: Efficient
   • Blog: 五分钟论文速读 | Spatially-Adaptive Feature Modulation for Efficient Image Super-Resolution
6. DLGSANet: Lightweight Dynamic Local and Global Self-Attention Networks for Image Super-Resolution
   • Paper: https://arxiv.org/abs/2301.02031
   • Code: https://neonleexiang.github.io/DLGSANet/
   • Keywords: Lightweight
   • Blog: 五分钟论文速读 | DLGSANet: Lightweight Dynamic Local and Global Self-Attention Networks for Image SR

盲超分

1. MetaF2N: Blind Image Super-Resolution by Learning Efficient Model Adaptation from Faces
   • Paper: https://arxiv.org/abs/2309.08113
   • Code: https://github.com/yinzhicun/MetaF2N
   • Keywords: Faces
2. Learning Correction Filter via Degradation-Adaptive Regression for Blind Single Image Super-Resolution
   • Paper: ICCV 2023 Open Access Repository
   • Code: https://github.com/edbca/DARSR
   • Keywords: Blind

Burst SR

1. Self-Supervised Burst Super-Resolution
   • Paper: ICCV 2023 Open Access Repository
   • Code:
   • Keywords: Self-Supervised
2. Towards Real-World Burst Image Super-Resolution: Benchmark and Method
   • Paper: ICCV 2023 Open Access Repository
   • Code: https://github.com/yjsunnn/FBANet
   • Keywords: Dataset

Reference-Based

1. LMR: A Large-Scale Multi-Reference Dataset for Reference-Based Super-Resolution
   • Paper: https://arxiv.org/abs/2303.04970添加链接描述
   • Code:
   • Keywords: Reference-Based

特殊场景

1. A Benchmark for Chinese-English Scene Text Image Super-Resolution
   • Paper: http://arxiv.org/abs/2308.03262
   • Code: https://github.com/mjq11302010044/Real-CE
   • Keywords: Scene text image super-resolution (STISR), dataset
2. Learning Non-Local Spatial-Angular Correlation for Light Field Image Super-Resolution
   • Paper: https://arxiv.org/abs/2302.08058
   • Code: https://github.com/ZhengyuLiang24/EPIT
   • Keywords: Light Field
3. Spherical Space Feature Decomposition for Guided Depth Map Super-Resolution
   • Paper: https://arxiv.org/abs/2303.08942
   • Code: https://github.com/Zhaozixiang1228/GDSR-SSDNet
   • Keywords: Depth Map

遥感

1. HSR-Diff: Hyperspectral Image Super-Resolution via Conditional Diffusion Models
   • Paper: ICCV 2023 Open Access Repository
   • Keywords: Hyperspectral, diffusion
2. ESSAformer: Efficient Transformer for Hyperspectral Image Super-resolution
   • Paper: https://arxiv.org/abs/2307.14010
   • Code: https://github.com/Rexzhan/ESSAformer
   • Keywords: Efficient, Transformer

医学

1. Rethinking Multi-Contrast MRI Super-Resolution: Rectangle-Window Cross-Attention Transformer and Arbitrary-Scale Upsampling
   • Paper: ICCV 2023 Open Access Repository
   • Keywords: MRI, arbitrary-scale
2. Decomposition-Based Variational Network for Multi-Contrast MRI Super-Resolution and Reconstruction
   • Paper: ICCV 2023 Open Access Repository
   • Code: https://github.com/lpcccc-cv/MC-VarNet
   • Keywords: MRI
3. CuNeRF: Cube-Based Neural Radiance Field for Zero-Shot Medical Image Arbitrary-Scale Super Resolution
   • Paper: https://arxiv.org/abs/2303.16242
   • Code: https://github.com/NarcissusEx/CuNeRF
   • Keywords: Arbitrary-Scale

视频超分

1. MoTIF: Learning Motion Trajectories with Local Implicit Neural Functions for Continuous Space-Time Video Super-Resolution
   • Paper: https://arxiv.org/abs/2307.07988
   • Code: https://github.com/sichun233746/MoTIF
   • Keywords: Continuous Space-Time
2. Learning Data-Driven Vector-Quantized Degradation Model for Animation Video Super-Resolution
   • Paper: https://arxiv.org/abs/2303.09826
   • Code: https://github.com/researchmm/VQD-SR
   • Keywords: Real-world
3. Multi-Frequency Representation Enhancement with Privilege Information for Video Super-Resolution
   • Paper: ICCV 2023 Open Access Repository
   • Code:
   • Keywords: Video

总结

1. SISR 领域中，接收的文章大部分都基于Transformer结构，展现出蹭热点的重要性。
2. 接收文章中提出了多个新数据集，集中在某些特殊场景（如burst，Scene text，Reference-Based等），有利于SR领域的进一步发展，也挖下了坑。
3. 轻量化超分文章比例大，展现出更偏实用化的研究趋势，在医学和高光谱等领域也涌现了相关研究。

参考资料

1. ICCV 2023 papers
2. ICCV 2023 超分辨率（Super-Resolution）论文汇总