Towards content-related features for parametric video quality prediction of IPTV services

Abstract

This paper investigates video content-related features, such as measures of spatio-temporal complexity, for inclusion into parametric video quality models. Our goal is to find a parametric content description that correlates with perceived video quality. In the course of the development of a parametric IPTV video quality prediction model (T-V-model), a large number of subjective tests have been conducted for standard definition and high definition video with different types of content. As expected from previous studies, we observed content dependencies that were different for different types of degradations. As descriptors of the content, we employ spatio-temporal related information obtained either before encoding and from the decoder or obtained from the decoder only. We compare those two approaches and explore their application to a reduced- or no-reference parametric model. An outlook highlights future steps for integrating the spatio- temporal features into the parametric model.