Guidelines to Implement Video Dataset to Assess Video Quality

Abstract

This article presents a novel video dataset featuring subjective evaluations, for analyzing performance of models predicting video quality of experience. Dataset Videos were distorted by modifying three quality of service parameters upon transmission over an emulated IP network environment. We obtained 60 distorted videos from 4 original source videos. The article presents criteria for selection of reference videos, the parameters for the subjective evaluation experiment, and the statistical treatment of results. Results evidence the impact of quality of service parameters in the observer-perceived quality. The dataset is publicly available upon request.