The way in which professional music is distributed and consumed has changed dramatically over the last 10 years. For this transitional period, the three key concepts that stand-out are Napster, iPod and Digital Rights Management (DRM). Currently, we have arrived at a stable situation where most of the digital audio distribution is controlled by a single retailer, and digital music is no longer encumbered by DRM. However, it is unclear that the distribution and consumption of professional digital video will follow the path of digital music. It might very well be that the future of digital video will include a strong DRM component. Why this might be the case, what form distribution of digital video will take, and why the inclusion of DRM might be less controversial than feared, will be the topic of this talk.

The teleconferencing industry newsletter Wainhouse Report defines Telepresence as "a videoconferencing experience that creates the illusion that the remote participants are in the same room with you." Today Telepresence is embodied in the marketplace by solutions such as HP Halo and Cisco Telepresence, dedicated conference rooms sporting built-in furniture and life-sized high-definition video, costing hundreds of thousands of dollars per room. In the future, Telepresence systems will be more diverse, enabling connections between not only meeting rooms but also offices, hotel rooms, vehicles, and even large unstructured spaces such as conference halls and stadiums. Mixed reality as well as ubiquitous computing - including robotics - will play key roles, because these systems will not only need to immerse the participants in a common world, but will also need to empower the participants in ways that are better than being physically present. In this talk, I will take you on a tour of various component technologies as well as experiences that are being developed in Microsoft Research for the future of Telepresence. Along the way will be evident many opportunities for advances in multimedia signal processing.

The audio quality of voice connections has remained virtually unchanged for more than 100 years. In most cases the audio bandwidth is still constrained to 3.5 kHz and nobody should expect to recognize, by listening to the sound, what is going on in the background of a call. With IP connections being used more and more for voice communication several attempts are now made to improve the situation. Some propose to considerably increase the audio bandwidth while others go as far as to promote communication in "CD-Quality" which could even include stereo or multi channel audio to fully transmit the accoustical image of the background of the speaker. What are the benefits to the user and what does it take to implement such services, as far as the audio components are concerned? This talk will try to give an overview about various systems proposed and what difference they can provide in user experience.

A typical 3DTV chain has capture, representation, compression, transmission, display interface and display stages. Each stage has its own specific nature and problems. And there are many alternative technologies for implementing each of these functional units. Signal processing tools play an important role in each such stage. The capture unit deals with difficult video data fusing problems. The post capture signal processing needs may range from nil in simplest 3DTV operations to demanding time-varying 3D model generation in sophisticated ones. Coding and compression of 3DTV video has its own specific nature and solutions. Probably the most complicated and demanding signal processing is at the display interface stage since 3D displays are quite different than 2D displays, and furthermore, since 3D displays come in many different forms. There are signal processing needs even within the camera and displays units. Among all different 3D modes, true 3D versions whichtarget physical duplication of information carrying light, such asholography and integral imaging, have their own rich signal processing needs. The signal processing problems associated especially with holographic 3DTV are unique and by far more demanding, and therefore, has the potential to trigger a new line of sophisticated signal processing techniques and associated mathematics.

Interactive poly-artistic works is a type of expression becoming increasingly common nowadays. Consequently, users, specta(c)tors, expect more and more to play an active part in these works. Such creations always require the use of a wide range of techno-logies (3D video and audio display, video and audio synthesis, body trackingÿ), and a large number of computer environments, software and frameworks have been created to fulfill these needs. However, despite this important profusion in terms of technical tools, several issues remain unsolved when realizing such artistic works. First, in the context of collaborative arts, existing frameworks do not provide means for con-ceptualizing art pie-ces for contributors coming from different artistic areas (composition, choreography, video, 3D graphicsÿ). Second, establishing communications between software or hardware components is often complicated. Finally, the communication process and its language have to be redefined from scratch for each new realization. We will introduce ConceptMove which is a unified paradigm for describing interactive poly-artistic works. In the second part of this talk we will focus on Interactive Storytelling, which can be regarded as a new genre, deriving both from interactive media such as video games and from narrative media such as cinema or litterature. Whatever degree of interactivity, free-dom, and non-linearity might be provided, the role that the interactor is assigned to play always has to remain inside the boundaries thus defined by the author, and which convey the essence of the work itself. This brings an extra level of complexity for writers, when tools at their disposal remain limited compared to technological evolutions.

Image based rendering (IBR) is a promising way to produce arbitrary views of a scene using images instead of object models. In IBR, new views are rendered by interpolating available nearby images. The plenoptic function, which describes the light intensity passing through every viewpoint in every directions and at all times, is a powerful tool to study the IBR problem. In fact, image based rendering can be seen as the problem of sampling and interpolating the plenoptic function.We therefore first briefly review some classical results on the spectral properties of the plenoptic function and then provide a closed-form expression for its bandwidth under the finite-field-of-view contraint. This naturally leads to an adaptive sampling strategy where the local geometrical complexity of the scene is used to adapt the sampling density of the plenoptic function. In this context, we also present an adaptive images-based-rendering algorithm based around an adaptive extraction of depth layers, where the rendering system automatically adapts the minimum number of depth layers according to the scene observed and to the spacing of the sample cameras. Finally, we discuss the problem of compressing the multiple images acquired for image-based rendering and present competitive centralized and distributed compression algorithms.This talk is based on work done with a number of collaborators, in particular, M. Brookes (ICL), C. Gilliam (ICL), A. Gelman (ICL), V. Velisavlievic (Deutsche Telekom) and J. Berent (Google inc.).