Perceptual video quality assessment in HTTP adaptive streaming

In this paper, we propose a quality metric for video transmitted over HTTP adaptive streaming (HAS). Perceived video quality in HAS strongly depends on the employed rate adaptation mechanisms, transmission conditions and buffer filling levels. Predicting perceived quality in HAS is an open problem, due to the lack of appropriateness of existing metrics. To overcome this problem, a quality metric based on extensive user evaluations is proposed in this paper. A thorough investigation is carried out, accounting for all possible types of video quality impairments occurring in HAS applications. The proposed metric correlates well with the perceived quality, allowing for optimizing the HAS descriptions and of the switching mechanism between them. It is also shown that the classical intrinsic quality metrics (PSNR, SSIM, VQM) fail to predict user-perceived video quality due to quality fluctuations inherent to HAS systems.