Maximizing Subjective Quality for Scalable Video Coding

With the upcoming scalable video coding algorithms as a solution to the problem of increasing heterogeneity in devices and networks, new challenges arise regarding the subjective quality towards the end user. Scalable video coding allows on-the-fly adaptation of the different aspects of visual quality, taking into account the expectations of the user. In this paper, we propose a machine learning approach that aims at predicting the preferences of a user regarding the different aspects of video quality (resolution, frame rate and image quality). This allows us to automatically select the best version available given a number of constraints imposed by the environment. Keywords—Multimedia, video quality, machine learning, scalable video coding, content negotiation