Evaluation of Adaptive Bitrate Selection Algorithms for MPEG DASH

Dynamic Adaptive Streaming over Hypertext Transfer protocol- HTTP (DASH) is a video streaming technique used for providing the end user with the best probable quality of claimed video sequence. HTTP based standard called MPEG DASH defines a way of notifying the user about different transport streams (TS) with different quality levels and gives information about the most suitable transmission path. MPEG DASH is used for encoding every video sequence into continuous segments of distinct alternatives: different resolutions, bit-rates, and quality levels. Based on present network conditions, an adaptive bitrate selection (ABR) algorithm at the client side selects the bitrate for each segment. In this paper, we evaluated three published ABR algorithms in a streaming client prototype based on the MPEG DASH standard. In the evaluation procedure, all algorithms were tested using three different video sequences from available MPEG DASH dataset. We determined the spatial and temporal information of all tested video sequences and their influence on PSNR (Peak Signal to Noise Ratio) values. Furthermore, video quality levels were analyzed in five testing scenarios in terms of different network conditions. Obtained results show that ABR algorithms that consider users experience and segment size are less sensitive to changing network conditions. The spatial activity has a high influence on the video quality and can be useful as a parameter in DASH dataset preparation in order to improve adaptation algorithms.