RealSR
收藏资源简介:
RealSR数据集由香港理工大学创建,包含595对高分辨率(HR)和低分辨率(LR)图像,用于真实世界的单图像超分辨率(SISR)研究。该数据集通过调整数码相机的焦距,捕捉同一场景的HR和LR图像对,并开发图像注册算法以逐步对齐不同分辨率的图像对。RealSR数据集涵盖室内外多种场景,旨在为SISR模型训练和评估提供真实世界数据,解决传统模拟数据集在实际应用中的局限性。
The RealSR dataset was created by The Hong Kong Polytechnic University, containing 595 pairs of high-resolution (HR) and low-resolution (LR) images for real-world single image super-resolution (SISR) research. This dataset captures HR and LR image pairs of the same scene by adjusting the focal length of digital cameras, and utilizes image registration algorithms to progressively align image pairs with different resolutions. Covering various indoor and outdoor scenarios, the RealSR dataset aims to provide real-world data for the training and evaluation of SISR models, so as to resolve the limitations of traditional simulated datasets in practical applications.

- 1Toward Real-World Single Image Super-Resolution: A New Benchmark and A New Model香港理工大学 · 2019年



