UGC-VIDEO: Perceptual Quality Assessment of User-Generated Videos
Yang Li, Shengbin Meng, Xinfeng Zhang, Shiqi Wang, Yue Wang, Siwei Ma
Abstract
Recent years have witnessed an ever-expanding volume of user-generated content (UGC) videos available on the Internet. Nevertheless, progress on perceptual quality assessment of UGC videos still remains quite limited. A distinguished characteristic of UGC videos in the complete video production and delivery chain is that they often undergo multiple compression stages before ultimately viewed and there does not exist the pristine source after they are uploaded to the hosting platform. To facilitate the UGC video quality assessment (VQA), we create a UGC video perceptual quality assessment database. It contains 50 source videos collected from TikTok with diverse content, along with multiple transcoded versions generated by different coding standards and quantization levels. Subjective quality assessment has been conducted to evaluate the video quality. Furthermore, we benchmark the database using existing quality assessment algorithms, and potential room is observed to further improve the accuracy of UGC video quality measures.