For a few years now, the WIC organises a yearly mid winter meeting. In these mid winter meetings, challenging topics in the field of information and communication theory are addressed. After turbo codes and speech recognition in the preceding years, this years subject was Object Oriented Coding for Multi Media applications. The meeting was held on Tuesday, January 19th in the Blue Hall at the Eindhoven University of Technology.
At 10:00, the chairman of the morning session, Jan Biemond of the Delft University of Technology opened the meeting. He welcomed the guests and thanked Andries Hekstra, Peter de With and Frans Willems for their help in the organisation. He also thanked the IEEE Benelux Chapter on Information Theory for their support. His opening talk included a short introduction into the topic of Object Oriented (OO) coding.
The first speaker of the morning session was Inald Lagendijk from the Delft University of Technology. He gave a clear introduction into the topic of OO coding; Due to the increasing demand for bandwidth by all kinds of applications, there is still an increasing need for more compression of the data. Until now, compression is usually performed on blocks of data, whereas it may be more efficient to use object oriented methods to represent the data. This makes it possible not only to compress, but also to manipulate the information in several ways. In order to compress efficiently, a priori information is required. He further addressed the topics which should be dealt with in more detail in later talks.
Next speaker was Rob Koenen of KPN Research. He explained about the concepts and the architecture of the MPEG4 standard; MPEG4 is a common technology which can be used for many types of services. A scene description is used both at the user and the server side. Since the system is not really real-time, the most important application areas are the Internet and Mobile Environments (both for broadcast and interactive services).
First speaker after lunch was Isabelle Corset of Philips Research. She made a lot of contributions to the standardisation of MPEG4. She explained the architecture of the encoder and told about the concept of sprites and how these are used in MPEG4 coding. In a few demos, the capabilities of MPEG were demonstrated. We saw several high quality video images in which she demonstrated the possibilities for adding and moving of objects.
OO coding requires segmentation of the original data, which Stef Desmet of ESAT addressed in his talk; The reasons for segmentation are the reuse of objects for editing purposes, scalability, graceful degradation and error concealment. A drawback is the extra cost for the segmentation mask. Segmentation is performed by using colour (texture), motion and connectivity characteristics in the (video)stream. Several different segmentation methods have been investigated, but a general solution is not yet available.
The last speaker of the day was Paul ten Hagen. The title of his talk was Facial Animation by Synthesis of Captured and Artificial Data. The goal of his research is to reproduce facial expressions on synthetic faces. Therefore, a lot of facial expressions, performed by actors, are recorded. This data is used to replicate the artificial expressions. The applications are very broad, e.g. telepresence, social user interfaces and applications in the film industry.
In the final talk of the day, the chairman of the afternoon session, Peter de With, concluded that MPEG4 is a shift from hardware to software. The very important topic of segmentation has to be worked on.
Concluding we have to say that we saw some very interesting talks. Ranging from the technical standards of MPEG4 via Segmentation (which still has to be further investigated) to some Applications of MPEG4. This introduction into Object Oriented Multi-Media applications has us hoping for the near future in which MPEG4 and its content-based interactivity will be reality.