Paradigms for Real-Time Video Communication and for Video Distribution

Sammanfattning: The use of new information technologies has drastically changed the way that we lead our lives. Communication technologies in particular have had a great impact on our day-to-day behavior. For example, it is now common to hear the voice and see the face of our loved-ones on another continents, or work with colleagues across the globe on a daily basis. With this change in behavior and the fast adoption of emerging technologies, new challenges in the telecommunications area are arising. This thesis is concerned with two such challenges: real-time video communication and video distribution. The latency constraint in real-time video communication is in essence incompatible with the uncertainty of best-effort networks, such as the Internet. The recent arrival of smart-phones has added another requirement to the application, in terms of the limited computational and battery power. The research community has invested a large amount of effort in developing techniques that allow a mobile sender to outsource video encoding complexity to an unconstrained receiver by means of a feedback channel. We question that approach with respect to real-time applications, arguing that long round-trip-times may render any feedback unusable at best, and costly in practice. We investigate the effect of channel round-trip-times on the popular distributed video coding setup, as well as on the traditional hybrid video coding architecture. Using a simple analytical framework, we propose the use of systems that adapt to the video content and the network in real- time. Our results show that substantial improvements in video quality can be achieved when the feedback channel is used correctly. The use of mobile devices has also a significant impact on the application of video distribution. In general, the multitude of devices that can be used to download and view video places new requirements on video distribution systems. The system must not only be able to scale to a large number of receivers in a bandwidth efficient manner, it must also support a wide range of network capacities and display capabilities. We address this problem by optimizing the set of rates that is used to provide video to receivers with heterogeneous requirements. Our approach is based on a favorable interpretation of the underlying mathematical problem, allowing the use of well-known quantization theoretic concepts. The resulting solution provides the possibility to design video distribution systems that adapt to changes in receiver characteristics online, with minimal delay.