video super-resolution quality assessment database (VSR-QAD)-1

Video super-resolution (SR) has important real world applications such as enhancing viewing experiences of legacy low-resolution videos on high resolution display devices. However, there are no visual quality assessment (VQA) models specifically designed for evaluating SR videos while such models are crucially important both for advancing video SR algorithms and for viewing quality assurance. Therefore, we establish a super-resolution video quality assessment database （VSR-QAD） for implementing super-resolution video quality assessment. Our VSR-QAD consists of 120 high quality high resolution reference videos. These videos were first spatially down-scaled by a factor of x2, x4, and x8, and then super-resolved back to their original resolutions by 10 representative SR algorithms to obtain 2,400 SR videos. Psychovisual experiments were carried out to acquire subjective quality labels for these SR videos. 190 subjects spending a total of approximately 400 hours to carry out both absolute rating and relative rating experiments. After aligning these relative and absolute scores and removing outliers, 2,260 SR videos are labeled with mean opinion scores (MOSs).视频超分辨率（Super-Resolution，SR）在现实场景中具备重要应用价值，例如可在高分辨率显示设备上优化老旧低分辨率视频的观看体验。然而，目前尚无专门针对SR视频评估的视觉质量评估（Visual Quality Assessment，VQA）模型，而这类模型对于推动视频SR算法迭代以及保障观看画质均具有关键意义。为此，我们构建了超分辨率视频质量评估数据库（VSR-QAD），用于开展视频超分辨率质量评估相关研究。该数据库包含120条高质量高分辨率参考视频：先将这些参考视频分别以x2、x4、x8的比例进行空间下采样，随后通过10种具有代表性的SR算法将下采样后的视频超分辨率恢复至原始分辨率，最终得到2400条SR视频。为获取这些SR视频的主观质量标签，我们开展了心理视觉实验：共有190名受试者参与实验，总耗时约400小时，完成了绝对评分与相对评分两类实验。在对相对评分与绝对评分进行对齐并剔除异常值后，最终为2260条SR视频标注了平均意见得分（Mean Opinion Score，MOS）。