G.1070 Model Extension at Full HD Resolution for VP9/HEVC Codec

Abstract

HD video streaming, which is gaining in popularity these days requires a large amount of bandwidth. This has resulted in the emergence of newer codecs like H.265/High Efficiency Video Coding (HEVC) and VP9 from Google. These codecs are supposed to provide an excellent video compression to quality ratio. ITU-T describes a standardised parametric model called the G.1070 Opinion Model, which estimates the Quality of Experience (QoE) of any multimedia content. The model estimates three parameters viz. the speech quality alone (Sq), the video quality alone (Vq) and the overall multimedia quality (M q) of the input video. However, it needs to be trained separately for different codecs, video formats and certain other parameters, which can be obtained by carrying out suitable subjective tests. Our contribution in this paper is threefold. First, we carry out a subjective test according to the Recommendation P.910 to estimate the video quality for VP9 codec. Second, for the first time we use the results obtained from the subjective test to find out a set of coefficients that enables us to extend the G.1070 model for VP9 codec at Full HD resolution. Third, we provide an answer as to which is the better codec from H.265/HEVC and VP9 by evaluating their performance against scores obtained from different standard objective tests like the G.1070 model, Video Quality Metric (VQM) model and the Peak Signal to Noise Ratio (PSNR) model.