The ever increasing availability of digital information in the form of images, videos, audio, and 3D models demands the design and development of systems enabling effective exploitation and management of such information.  This relates to how multimedia information can be coded, transmitted, and made available to users via high-speed networks, but also, and primarily, to how multimedia information can be stored and accessed. User access to multimedia—particularly visual—data is mainly based on high-level, semantic concepts, while automatic algorithms can extract only low-level features.  Bridging this so-called semantic gap is one of the most challenging problems in enabling effective access to information in multimedia form.

Addressing these issues requires the contribution of many research disciplines and poses new research challenges to researchers in this field.  Image/video analysis and processing, information retrieval, pattern recognition and computer vision, multimedia data modeling, multidimensional indexing, psychological modeling of user behavior, man-machine interaction, and information visualization, are not even all of the research fields that contribute to this new area of research.

IAPR's Technical Committee 12 (TC12) on Multimedia and Visual Information Systems promotes interaction among researchers working in modeling, design, and development of systems for the analysis, processing, description, and retrieval of multimedia,  particularly visual information, as well as the applications of these systems in challenging domains.  Activities of TC12 center round the topics of benchmarking and education.

Benchmarking:  Progress in such a complex and diverse field as multimedia and visual information systems can only be measured if various methods can be compared against one another using objective benchmarks.  This has been recognized widely in the field and many benchmarks are available.  Different benchmarks focus on different aspects of the problem.  The Amsterdam Library of Images focuses on retrieving images of objects when recorded under varying conditions.   For Content Based Image Retrieval,  BENCHATHLON has addressed many issues in benchmarking, including the search for edited or re-compressed images originating from the same source.  The previous TC12 chairs developed a benchmark focusing on semantic access to a picture collection.  For video, the TRECVID benchmark addresses semantic access to the data, also.  Clearly for video, techniques for indexing and search can make use of not only the visual, but also of the audio/speech channel.  The Princeton Shape benchmark has a large collection of 3D models.  TC12 aims to promote these benchmarks within the IAPR community and, in addition, will work on new benchmarks specifically focusing on methods that measure the effectiveness of the interactive retrieval process as a whole.

Education:  Due to the variety of disciplines forming the basis for techniques in Multimedia and Visual Information Systems, education in this field is still a challenge.  Often the topics are taught from, for example, a computer vision, machine learning, or visualization perspective.  An integrated view on the topic requires knowledge of each of the contributing disciplines.  Clearly having thorough knowledge of all these disciplines is unfeasible. Therefore, the TC12 will collect links and references to educational resources for both students and teachers, making it easier to study the problem in the genuine, multi-disciplinary perspective.  This will not be limited to papers and books, but will also include interactive materials and demonstrators that allow students to have interactive learning experiences.

We are taking the challenge to contribute to the progress of this exciting field, but of course can only do so together with you, our fellow researchers in the field. We invite you to visit the TC12 web site www.science.uva.nl/~worring/TC12 for more information, and to contribute ideas and resources.

Technical Committee Report: 

TC12 Multimedia and Visual

Information Systems