VLBV98 Panel 5: MPEG-7 Issues

VLBV 98: Oct 8-9, 1998, University of Illinois at Champaign/Urbana


M. Ibrahim Sezan  (panel statement of purpose):

The goal of this panel is to discuss the issues around the upcoming MPEG-7 standardization effort. MPEG-7 addresses standardization of common interface for describing multimedia information. Major applications range from multimedia database access to multimedia information filtering using the standardized descriptions. MPEG-7 is planned to reach International Standard status in year 2000.

MPEG-7 is the most recent activity of Moving Picture Experts Group (MPEG) who has just completed the Version 1.0 of the MPEG-4 standard. The preceding MPEG standards, MPEG-1, 2, and 4 have addressed coded representation of audiovisual information. In one sense, a coded representation may serve as a description and hence MPEG-7 and previous MPEGs have commonalties. On the other hand, there are also important differences between MPEG-7 and previous MPEGs in terms of technical tools involved and applications.

MPEG-7 aims at standardizing a core set of descriptors (D), i.e., quantitative measures of audiovisual features, description schemes (DS), i.e., structure of descriptors and their relationships, and a description definition language (DDL) to express description schemes. The focus of MPEG-7 is on audiovisual information, that may contain text, but not on text-only documents. MPEG-7 will embrace existing text descriptors as they are appropriate for text-based descriptions of audiovisual information. It is expected that the standard core set will facilitate those classes of applications that have wide-spread usage and will provide interoperability. For specific applications served by outside the core set, it will be possible to use the DDL to define a specific DS. It is not decided at this moment how such extensions will be registered or incorporated into future versions of the standard. Feature extraction or descriptor computation, search engines and database organization, are left outside the scope of the standard. The motivation is to allow different vendors to compete on these issues and also not to freeze these tools which are still among topics of active research and have promise for future improvement.

I think the differences between MPEG-7's scope and that of previous MPEGs is obvious. But this is not the only difference. MPEG-7 brings two previously separated groups, namely the "database community" and the "signal processing community", together. These two groups have often have different views and practices, such as dealing with semantic but qualitative versus low-level but quantitative features. Further, although need for MPEG-7 is well understood and appreciated, its emphasis on audiovisual information is still new in a world dominated by text databases. Hence, the MPEG-7 group is still spending a lot of time on defining the scope, requirements of the standard as well as evaluation of tools that will form the standard. Again, evaluation of a text database is relatively much better understood than evaluation of an audiovisual database, at least due to the fact that "ground truth" may not be easy to specify in case of audiovisual content. The members of the panel will touch various important MPEG-7 related issues. Among those that are of major importance are the following:

  1. Views from two separate groups (database and signal processing groups) and need for harmonization
  2. What is the balance between flexibility and compatibility (i.e., standardizing only a DDL versus standardizing the DDL and a core set of D and DS)
  3. What should the goal of this core standard set be to ensure wide-spread usage and early adoption of the standard?
  4. What if a certain descriptor requires standardization of the method of extracting the feature it represents? (Note that feature extraction currently is outside the scope of MPEG-7)
  5. What kind of descriptors and description schemes are useful for video?
  6. Is it possible to relate browsing with description schemes?
  7. Can we achieve real interoperability if we do not specify what results a search engine or filter agent should produce given a certain descriptor?
  8. How can we ensure speed and accuracy, both among the requirements of MPEG-7, if we leave search engines and database organization outside the scope of the standard?
  9. Importance of descriptor generation tools in making MPEG-7 a reality