direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

Free Viewpoint TV


Motivation & Project Description Given the fact of the Telekom AG interest in Internet TV (IPTV) one of the key issues is   giving a user a complete  freedom to  interact with  the broadcasted  video. The  main idea  is that  the  user can arbitrary choose a new viewpoint from which he wants  to view the scene and to be in full control of changing the viewpoint  anytime he wants. It is  possible to imagine plenty of  scenarios where this will be very useful feature: sport events,  movies and entertainment programs. The system which will provide  such   a  service  has  to   be  based  on   multi-camera  system  which  film   the  scene simulateniously. The 3D representation of the  scene has to be reconstructed from those synchronized multiple videos in real time and multicasted to the multiple users. Having a complete  textured 3D representation of the scene one is able to  move a  virtual camera  around the  scene in  any freely  chosen viewpoint.   Having the 3D information and  the texture on the client side the user can freely change his view point while the scene is dynamically changing, to freeze the scene and move around and eventually augment the scene by changing the surroundings. The following topics will be covered by this project:

  • 3D Reconstruction

    • Development of an advanced  algorithms for 3D reconstruction from multiple cameras
    • Enhance shape using photoconsistency and temporal coherence

  • Multiple view  video compression, coding and interpolation

    • Efficient coding of multi-view video signals
    • Edge-adaptive interpolation of lightfield in between multiple cameras

  • Camera callibration

    • Enable calibration from raw video of the network of cameras.

  • Interaction

    • Vision based intuitive display and control of viewpoint for user.


Expected Outcome: The project consists of research and technology parts. The research should result in scientific publications and patents, while technology part should produce a demonstrator for T-Galleries.

Time Frame: 

07/2007 – 07/2010
T-labs Team Members:
Rahul Swaminathan, Slobodan Ilic, Vladan Velisavljevic, Michael Rohs
Students:


Partners:
PERCEPTION, INRIA, Rhone-Alpes / HHI, Fraunhofer Institut, Berlin / GRASP Laboratory, University of Pennsylvania
Funding by:
Deutsche Telekom Laboratories
Publications:
  • E. Ekmekcioglu, V. Velisavljevic and S. T. Worrall, Content adaptive enhancement of multi-view depth maps for free viewpoint video, IEEE Journal of Selected Topics in Signal Processing, to appear April 2011
  • V. Velisavljevic, Method for encoding texture data of free viewpoint television signals, corresponding method for decoding and texture encoder and decoder, EPO Patent pending, filed March 2010
  • M. Morbee, V. Velisavljevic, M. Mrak, W. Philips, Scalable feature-based video retrieval for mobile devices, International Conference on Internet Multimedia Computing and Service (ICIMCS), November 2009
  • E. Ekmekcioglu, V. Velisavljevic, S. T. Worrall, Efficient edge, motion and depth-range adaptive processing for enhancement of multi-view depth map sequences, IEEE International Conference on Image Processing (ICIP), November 2009
  • E. Ekmekcioglu, V. Velisavljevic, S. T. Worrall, Edge and motion-adaptive median filtering for multi-view depth map enhancement, Picture Coding Symposium (PCS), May 2009
  • V. Velisavljevic, Edge-preservation resolution enhancement with oriented wavelets, IEEE International Conference on Image Processing (ICIP), October 2008
  • V. Velisavljevic, M. Mrak, A. Kondoz, Application of directionlets in video coding, Visual Information Engineering, July 2008
  • V. Velisavljevic, R. Coquoz, Image interpolation with directionlets, ICASSP, April 2008

Zusatzinformationen / Extras

Direktzugang

Schnellnavigation zur Seite über Nummerneingabe