Subjective Study of Perceptual Quality for Micro-Video Applications

Past few years have witnessed the surprising popularization of micro-videos. Various micro-video applications have occupied a dominant portion in the mobile application market. To enhance user experience, it is crucial to explore the perceptual quality of micro-videos. In this paper, we establish a new subjective quality assessment database for micro-videos. The database consists of 121 mobile user-captured videos and mean opinion scores (MOS) generated by 2541 subjective rating scores from 21 naive subjects. The videos are chosen to be representative for micro-videos, including different capture devices, composition styles, orientations, content categories, and a wide range of authentic and complex distortions. The subjective test is conducted on mobile phones, simulating the realistic viewing condition of micro-video applications. With this database, we also tested the performance of some existing no-reference (NR) video quality assessment models to verify their ability of handling the realistic distortions in micro-videos. The proposed database will be made available to the researchers so that it can help the future work on the study of perceptual quality of micro-video applications.