Performance of four subjective video quality assessment protocols and impact of different rating pre-processing and analysis methods

Abstract

Quantification of imaging system performance and training/validation of objective quality assessment (QA) algorithms both require accurate estimation of subjective video and image quality. The International Telecommunication Union (ITU) offers various recommendations for subjective testing protocols, including paired comparison, double stimulus, and single stimulus methods. While many studies have compared the performance of subjective QA protocols, different methods for pre-processing and analyzing rating scores are not well studied. Many studies often pre-process or transform the raw ratings generated from a given protocol, even though some of these techniques are not standardized. In addition, ratings may be analyzed using different statistical methods. Finally, the discrimination performance of protocols may differ depending on the magnitude of the quality differences in the test dataset. We have not found any studies in the literature that have conducted a comprehensive analysis of these effects, especially in video QA.

Publication
IEEE Journal of Selected Topics in Signal Processing