MediaSemantics - Chitra Dorai, IBM Research

MediaSemantics

The MediaSemantics project in the Computing Utilities (aka e-Utilities) Department is developing automatic content identification and high-level concept-oriented description systems to deliver rich video annotation, search and browsing services. It takes an distinctive approach by focusing not merely on the representation of perceived video content, but also on the semantics and the emotional impact of the content seen and remembered.

Our recent effort includes pioneering a new research approach in media analysis called the Computational Media Aesthetics to address the problem of semantic gap in automatic content annotation systems. Computational media aesthetics is defined as the algorithmic study of a variety of image and aural elements in media, founded on their patterns of use in film grammar, and the computational analysis of the principles that have emerged underlying their manipulation, individually or jointly, in the creative art of clarifying, intensifying, and interpreting some event for the audience.

This media production-guided semantic analysis approach has been well received and followed in the research community. See Honors for the latest awards and recognition for this work. This is joint work with Prof. Svetha Venkatesh and our graduate students.

Back Last modified: Wed Dec 13 15:56:11 2000