A Subjective Quality Assessment Database for Mobile Video Coding

As mobile devices and network popularize significantly, increasing number of videos are captured and viewed with mobile devices. In this paper, we introduce a new subjective quality database for mobile videos. The database contains 128 distorted videos generated from 32 source videos captured with different mobile phones. Among the source videos, 28 out of 32 are in portrait mode that occupies a significant portion in current mobile video applications. Distortions are introduced by compressing the videos in different bitrates. The subjective quality rating tests are conducted on mobile devices, involving 23 naive subjects and finally resulting in 2816 quality rating samples and corresponding mean opinion scores (MOSs) for all the 128 distorted videos. Furthermore, we also evaluate some objective quality assessment models on the database to verify their capability on predicting the quality of mobile videos. The proposed database will be made public to facilitate future research on mobile video quality assessment.

I. Introduction

As mobile devices and mobile network popularize rapidly, increasing number of videos are captured and viewed via mobile devices. Given that mobile videos play an increasingly important role in video applications nowadays, the study on mobile video quality is of much significance for providing better quality of experience.