第一次来请先看这篇文章:【超分辨率(Super-Resolution)】关于【超分辨率重建】专栏的相关说明,包含专栏简介、专栏亮点、适配人群、相关说明、阅读顺序、超分理解、实现流程、研究方向、论文代码数据集汇总等)
文章目录
- 前言
- Abstract
- 1. Introduction
- 2. Related Work
- 3. Real-world SISR Dataset
- 4. Laplacian Pyramid based Kernel Prediction Network
- 5. Experiments
- 6. Conclusion
- 7. Supplementary Material
-
- 7.1. Sample images of the RealSR dataset
- 7.2. The details of our network architecture
- 7.3. More visual results by SISR models trained on simulated SISR datasets and our RealSR dataset
- 7.4. The computational cost of competing models and their visual results when trained on our RealSR dataset
- 7.5. More super-resolved results on images outside our dataset
前言
论文题目:Toward Real-World Single Image Super-Resolution: A New Benchmark and A New Model —— 迈向真实世界的单图像超分辨率:一个新的基准和一个新的模型
论文地址:Toward Real-World Single Image Super-Resolution: A New Benchmark and A New Model
论文源码:https://github.com/csjcai/RealSR
ICCV 2019!重名的RealSR!
Abstract
现有的基于学习的单图像超分辨率 (SISR) 方法大多在模拟数据集上进行训练和评估,其中低分辨率 (LR) 图像是通过对高分辨率 (HR) 对应物应用简单且统一的退化(即双三次下采样)来生成的。然而,现实世界 LR 图像中的退化要复杂得多。因此,在模拟数据上训练的SISR模型在应用于实际场