The MultiMedia Project

Unified Methods and Algorithms for Joint Storage, Orgnization and Labeling of Image, Voice and Text Data in Multimedia Applications on the Information Highway

Goals: Development of methods and algorithms, which will allow clever management of complicated information systems on the Information Highway

Participants: Prof. Amir Averbuch, Prof. Arnon Cohen, Prof. Itshak Dinstein, Prof. Ehud Gudes, Prof. David Malah

Abstract

The information highway is expanding very rapidly, and thousands of new users are using it every day, accessing huge collections of multimedia data. Efficient tools for storage, search and navigation are essential. This project deals with development of algorithms for compression and indexing (labeling) of multimedia data consisting of imagery, speech, signals and text. Advanced novel image compression techniques are proposed. These techniques are intended to enable very deep image compression (e.g. less than 0.1 bit per pixel for still images). Compression of the voice, signal and text data will be handled in a conventional manner. Indexing and retrieval of the data will be achieved with parametric as well as with non-parametric techniques. The system is developed within Object Oriented framework (Java) and an Object Relational Data Base. This will enable handling of content related data items that are very different in terms of their structures and sizes. The item indexing within the database is dynamic. New tags, indices, and labels will be attached to data items as a result of users queries.

The proposed research is performed by five groups located at three universities. Three groups are with Ben-Gurion University, an image processing group headed by Prof. Dinstein, a speech and signal processing group headed by Prof. Cohen, and the database group headed by Prof. Gudes. A second image processing group is headed by Prof. Averbuch of Tel-Aviv University, and a third image processing group is headed by Prof. Malah of the Technion.

The first application of this project is a prototype of a patient file in a typical Oncology department in a large hospital. Such a file may contain many images of various types, such as CT, X-Ray, MRI, Ultra-Sound, etc., voice files explaining these images or other clinical information, and textual information such as results of Blood and Chemical tests. Such a system can help in accumulating all the relevant information in a single place and can help enormously a physician that has to decide on treatment based on all of this information, as well as enabling comparison with historical information and other patients information. Obviously, this information could be transferred over the Internet for consultations. Clearly, such large quantities of information (a single patient may require tens of images!) demands efficient compression with the challenge of not losing the important features. It can directly use the results of the proposed research.

The proposed research project has profound potential benefits, both in commercial terms, as well as in terms of the general public benefit. In addition, it brings together researchers of various fields (image, speech, databases) that usually work independently. This cooperation has a high potential for mutual stimulation and creativity.