The Media department at Simula branches out into several research disciplines - from video encoding and 3D rendering algorithms to interactive multimedia performance and communication protocols. Interactive time-dependent applications are used daily by large user masses of the Internet, being a great motivation for our research on reducing Internet transport latency and improving the performance of latency-sensitive services. The department has a holistic approach to media dissemination: apart from analysing data, improving media efficiency and distribution algorithms, we contribute to the field of user perception. Exploring perceptivity can assist in tailoring the performance to fit increasingly precise parameters for human comprehension. Our psychologists study human reactions to distributed data, signal delay and data lag. The main goals of Media department are reducing the costs and speed of media distribution, increasing the number of users and optimizing the perceived service quality.
Brief Overview of Activities
Video Delivery Systems
MPG covers a range of video delivery systems ranging from traditional video streaming, video search to 3D representation of content and delivery of free-view video. In a fairly mature part of the field, we address questions of efficient delivery and best user experience when using the adaptive segmented HTTP streaming that is currently favoured by industry. In a very young part of the field, we investigate means of combining video streams from sparse camera arrays that monitor large space to generate 3D scenes in real-time and deliver them to viewers.
Perceptual Video Quality Assessments
To make the right decision in video coding, adaptation during transport and rendering on users' screens, it is necessary to understand how these decisions affect the people who watch the result. The delivery of perfect quality is always inhibited by limited resources, such as recording devices, computing power, network bandwidth and display devices. Adapting to available resources involves a choice of, for example, compression method, target bandwidth, quantization factor, resolution or frame rates of videos. Whether decisions are good is known to depend on external conditions as well as content and user intent, but how to make good decisions is a wide-open research field, and one that MPG pursues.
Distributed Data Processing
Multimedia data, and especially the current trend towards 3D scene creation and rendering, generates large data volumes that must be processed in a timely manner. With growing complexity and an increased user expectation to use complex media, the computational demand for multimedia data processing grows. The latest hardware architectures use an increasing number of heterogenous multi-core processors to cope with the demand for computing power. Developers find it much harder to develop application for such architecture than for homogeneous sequential ones, and MPG investigates how multimedia applications, which differ from other workloads by their need for deadline-driven scheduling, their ability to trade speed for quality and their cyclic nature, can be developed for and deployed on such architectures.
- P2G - The P2G framework is a distributed execution system for processing of continuous multimedia wordloads on heterogeneous architectures.
Existing data communication protocols are designed for best-effort networks, and transport protocols today are optimized for high-bandwidth applications like plain file transfers and web-browsing. Interactive distributed multimedia applications however have a different requirement, the delivery of usually small amounts of data within consistently short times. Even though interactive applications that have this requirement comprise a large number of connections in the current Internet, they are hardly researched at all and existing solutions fail. MPG investigates new protocols and mechanisms that address this challenge and that realistically deployable in the Internet.