Quality Adaptive Video Caching and Transport in Networks

The transport of video flows is a demanding task in distributed multimedia systems based on complex, heterogeneous networks. Currently, there are two major approaches for supporting video data transport in such networks, viz. resource reservation schemes and adaptation techniques residing in the end-system applications that provide or utilize the videos. The proposed project pursues an alternative approach: it systematically explores how video adaptation (re. quality and resource consumption) can-be delegated to, or supported by, the transport infrastructure, i.e., by network nodes (such as routers) and proxy caches located in the network. Specifically, it is proposed that active network nodes (routers) change video flows during transport in that they perform media scaling (and thus, quality variation) operations when required. Examples of such cases are: fluctuating network QoS, in particular congestion situations; delivery of video flows along heterogeneous paths or to end systems with different playout capabilities; and multicast transfers. The network is best positioned to quickly and gracefully react to such situations. Moreover, it is proposed that active proxy caches exploit video adaptation opportunities to provide novel cache management and flexible buffering techniques. An example is that the quality (and thus storage requirements) of a video clip is reduced (in several iterations) before it is entirely replaced from the cache store. The project shall make use of and combine three recent developments in multimedia communication and networking research: (1) a variety of video scaling options offered by modem coding standards like MPEG-4, both on a low level (filtering techniques, layered coding, fine-granular scalability) and on a higher, content-based level (e.g., priorities of video objects); (2) descriptive data about the videos and their adaptation options (akin to MPEG-7 variation descriptors), to be stored as part of the videos` meta-data and used to govern the adaptation process in the network; and (3) active network concepts and a prototype active network software system. The project shall proceed and provide results as follows: adopt and develop, if not readily available, MPEG-4 based video coding and scaling software; identify, model, store as meta-data, and utilize relevant MPEG-7 based video adaptation descriptors (jointly with the project proposed by Kosch/Böszörmenyi); design, implement, and evaluate the basic mechanisms (not policies) for video adaptation in an active network; investigate and evaluate novel adaptive proxy cache management and buffering algorithms; integrate the mechanisms into an appropriate protocol architecture (RTSP and RTP/RTCP); and demonstrate and analyze the mechanisms and algorithms in a suitable larger multimedia scenario.

The last decade has brought about an enormous growth of digital multimedia content: graph-ics, images, audio, and video. Multimedia contents, applications, and user interfaces will in-creasingly determine how computing, communication, and entertainment will look like in the future and be used in everyday life. Digital media are produced and offered in diverse media formats, for instance MP3 for music or MPEG-2 for Digital TV. At the same time, there is a growing diversity of devices via which access to multimedia content is desired, ranging from High Definition TV to mobile phones capable of playing videos. The same holds for the networks which are being used to transmit multimedia contents, ranging from satellite networks to Wireless LANs and UMTS. This situation raises a fundamental problem: how to transmit the plethora of digital multime-dia contents over the heterogeneous network infrastructure, which is even subject to dy-namically changing traffic and load conditions, to the many different end user devices, and how to play or display the content in an optimal way. This should be enabled at minimum costs; for instance, generating and storing an extra version of an audio-visual content for any media format / network / device combination conceivable, is certainly not an option. This project systematically investigated, in the demanding domain of digital videos, how ad-aptation of the videos could be utilized to solve this problem. As a major result in the area of multimedia transport, a video streaming system was built that can tailor video contents to the end user`s device capabilities and that is highly adaptive w.r.t. dynamically fluctuating network conditions. Videos are adapted in multiple dimensions: tem-porally (change of video frame rate), spatially (change of the size of the images, which is however compensated by the video player), and in quality (higher compression ratio). Further, a video proxy server was created and evaluated that can, for instance, buffer video contents near the consumer (cache functionality) and tailor the video contents to the device capabilities (adaptation functionality). The proxy module serves as a mediator (gateway) among the device properties and the available video variants and is able to generate new version of the content in real time, if required. Both sub-projects intensely used modern standards for video and audio compression (MPEG-4), metadata (MPEG- 7 and MPEG-21), and for communication (Internet protocols like RTP). Examples of the many research questions addressed in project are, how to ac-quire up-to-date information about network conditions (for fast reaction in the streaming sys-tem) or how to find optimal adaptation decisions in the proxy server, using a quality and cost model. A rich software base called "The Video ToolKit (ViTooKi)" for adaptation and transport was built and made available to the open source community (vitooki.sourceforge.net).