DRealSR

DRealSR建立了具有不同现实世界退化过程的超分辨率 (SR) 基准，从而减轻了常规模拟图像退化的局限性。
它是从五个DSLR相机在自然场景中收集的，并覆盖室内和室外场景，避免移动物体，例如广告海报，植物，办公室，建筑物。训练图像被裁剪成380 × 380、272 × 272和192 × 192补丁，产生31,970补丁。

DRealSR establishes a super-resolution (SR) benchmark with diverse real-world degradation processes, thereby alleviating the limitations of conventional simulated image degradation. This benchmark is collected using five DSLR cameras in natural scenes, covering both indoor and outdoor scenarios, while excluding moving objects such as advertising posters, plants, offices, and buildings. The training images are cropped into patches of 380 × 380, 272 × 272, and 192 × 192, resulting in a total of 31,970 patches.

搜集汇总

背景与挑战

背景概述

DRealSR是一个超分辨率基准数据集，包含从自然场景中收集的图像，覆盖室内外场景，避免移动物体，训练图像被裁剪成不同大小的补丁，共31,970个。该数据集由多所大学于2020年发布，旨在解决常规模拟图像退化的局限性。