MISTRAL - Managing Internet video STReaming through Adaptation and Learning

About

The goal of MISTRAL is to investigate the next generation HTTP (Hypertext Transfer Protocol) adaptive streaming architecture, protocols, and algorithms for high quality media streaming over IP networks in general, and the Internet in particular.

Both extensions to the current 3rd generation of AVC-based (Advanced Video Coding) HTTP adaptive streaming, and a 4th generation SVC-based (Scalable Video Coding) HTTP cognitive streaming will be researched, targeting disruptive innovations in the video streaming domain.

The technical requirements for a next generation HTTP streaming will be determined based on three complementary and equally challenging use cases:

  1. Multi-site collaboration
  2. Internet-based media production and delivery
  3. Web television

Next, in-depth research will be done along 4 dimensions:

  • Optimal video encoding and temporal segmentation for H.264/AVC and H.264/SVC in the context of HTTP-based delivery, optimization of SVC encoding for the same context, and study the efficient encoding of synthetic content (combining images, video, data, and graphics).
  • Architectural extensions, features, and interaction with network level functionality (e.g. appropriate linking with network traffic engineering, proactive video segment downloads) that can improve HTTP-based delivery of video segments.
  • HTTP adaptive streaming quality monitoring mechanisms and a novel perceptual video quality metric, based on detailed experiments and user experience studies for the three defined use cases.
  • Exploration of self-learning mechanisms that allow the streaming system to gradually improve over time, these mechanisms will be studied for the encoding side, the cognitive network elements and the client-side decisions.

Methodology

This project is multidisciplinary since it requires the combined expertise in various domains:

  • advanced networked system design
  • network management
  • video encoding
  • advanced optimization algorithm design
  • user perception modeling
  • subjective and objective video quality assessment

Research results will be thoroughly analyzed through a variety of simulation and emulation experiments, and extensive subjective testing in a two phase approach. Finally, the obtained results will be integrated in three Proof-of-Concept (PoC) implementations (one for each use case) and detailed validation and evaluation will be performed on each of the PoCs.

The consortium is composed of Flemish leading stakeholders in the domain of video delivery:

  • VRT: content provider
  • Belgacom: network operator and IPTV service provider
  • Alcatel-Lucent Bell Labs: network equipment provider
  • Barco: high-end user equipment and networked application provider

The industry consortium is convinced of the high-added value of jointly investigating in detail the novel added-value of adaptive video streaming for their future products.

Five iMinds research groups are involved, one of them is MICT.

Duration of the project

The project runs from 01/01/2013 - 31/12/2014.

Staff involved

Financed by