News item
8 maart 2010
15th MuMe-Community Workmeeting report
In the afternoon of February 25, the 15th work meeting of the MuMe-community was held in Leuven.
33 members joined for an interesting program. The gathering was organised in a brand new meeting room to provide enough space to allow that the presentation of Philippe Bekaert also served as an internal seminar for an about as large group of interested imec people. It also allowed to stream this presentation to the imec center in Eindhoven.
Philippe Bekaert form the Expertise center for Digital Media of the Hasselt university kicked off with an overview of the pioneering work in his group on combining video into a real 3D experience. After a quick review of previous approaches towards 3D representations, Philippe presented their system based on multi-camera recording, processing of these streams with advanced innovative algorithms and on-line generation of views, of which the direction and zoom factor can be selected by the user. These features were demonstrated on a walkthrough experience in Bokrijk and immersive video position experiences on a musical event in Brussels and an American football game. After discussing a number of algorithmic examples and pointing out that real progress is being made, Philippe concluded that still a lot of interesting work must be done.
Johan De Geyter from imec, stood in for Francesco Pessolano to present the NVision program. This program aims at realising efficient building blocks and systems for advanced image applications, exploring the use of combinations of advances in technology at different levels. Johan presented the experience build up on the efficient implementation of video coding and other vision applications like de-noising, stereo matching and viewpoint interpolation as well as the different platforms used (multiprocessor ASIC, FPGA-board, ..). Johan finally pointed out the topics of focus in ongoing work, e.g. in the field of hyper spectral imaging applications.
Switching to the audio domain, Jean-Pierre Martens from the University Ghent discussed the status and recent advancements on “Audio Indexing for Information Retrieval”. Jean-Pierre set the scene by presenting a number of applications and their main requirements. He then discussed the subsequent steps needed to generate the necessary meta-data. These steps were highlighted by results on a set of test data from a European contest and a live demo of the segmentation and clustering steps in their system. To conclude Jean-Pierre presented the use of audio indexing systems to realise high quality subtitling of TV programs.
In the last presentation, Matthias De Geyter, from the VRT-Medialab, presented the fundamentals, technologies and products of an end-to-end 3D TV chain. After an historic background starting as early as Euclides, Matthias discussed the fundamental principles and problems of stereoscopy, the different steps in an end-to-end 3D system from camera to display and the challenges in each of the steps. In the last part of the presentation he presented some recent advanced technologies, like autostereoscopic 3D and beyond as well as 3D audio that is also important to give the users a realistic 3D experience.
The meeting was concluded by an informal gathering where the attendees had some lively discussions on the technologies presented, improvements to be investigated and topics for the next meetings.
Based on the feedback at and after the meeting and on the evaluation forms (scores all above 8) the presentations really met the expectations of the attendees.
Our sincere thanks to the speakers for their enthusiasm, the well selected material and demos and their really interesting talks.
The next meeting is planned on May 27 in the afternoon. More info on the agenda will be posted soon.
