Introduction

Abstract

Video content accounted for almost two thirds of the world’s consumer internet traffic in 2014, and is predicted to account for 80% by 2020; by the end of this decade, it is expected that almost one million minutes of video content will cross global IP networks every second Cisco (Cisco visual networking index: forecast and methodology, 2015–2020, 2016) [1]. According to Sandvine’s 2016 “Global Internet Phenomena Report” Sandvine (Global internet phenomena, 2016) [2], video streaming accounts for over 60% of peak-hour broadband internet traffic consumption in North America, with Netflix (35%) and YouTube (18%) being the main contributors. In a video streaming scenario, a variety of users with different resources in terms of screen size, resolution, processing power, and network bandwidth, are accessing the same video content, as illustrated in Fig. 1.1. Currently, the heterogeneous requirements of web streaming are met by storing hundreds of copies of the same video on the server Gigaom (To stream everywhere, netflix encodes each Movie 120 times, 2012) [3]. Clearly, there exists a lot of redundancy between the different copies; the reason for this “wasteful” storage is that existing video coding standards (e.g., H.264/AVC Wiegand, Sullivan, Bjøntegaard, Luthra (IEEE Trans. Circuit Syst. Video Technol. 13(7):560–576, 2003) [4] and HEVC Sullivan, Ohm, Han, Wiegand (IEEE Trans. Circuit Syst. Video Technol. 22(12):1649–1668 2012) [5]) are optimized for a predefined set of network and decoder constraints.