COLLABORATIVE RESEARCH: CONTENT BASED QUERIES IN IMAGE DATABASES

Ramesh Jain

Department of Electrical and Computer Engineering
University of California, San Diego

Contact Information

Ramesh Jain
Dept. of Electrical & Computer Engr.
University of California, San Diego
9500 Gilman Dr.
La Jolla, CA 92093-0407
Phone: (619) 534 8639
FAX: (619) 534 1004
e-mail: jain@ece.ucsd.edu

WWW PAGE

Image Database info

Keywords

Search by content, emergent semantics, user interfaces, human machine interaction, similarity.

Project Award Information

Award Number: IRI-9610518
Duration: 9/1/97 - 8/31/98
Title: Content-Based queries in Image Databases

Project Summary

In this project, we study new ways of interaction between a user and an image database. Existing databases are based on relatively low-level image features, like color histogram, color regions, edges. These features can be extracted in an efficient and reliable way from a variety of images, but their semantic content is quite far from what the user wants. The user thinks of images in terms of objects, situations, and other high level constructs.
Rather than trying to extract semantically meaningful features from the images (a task that, in the general case, is too difficult for current and foreseeable techniques), we rely on the interaction between the user and the database to let the semantic emerge. The user explores
Goals, Objectives, and Targeted Activities
The project studies new modes of interaction between users and image databases. Our goals for the next year are the following.

Evaluation.Indexing.Meta Dataentire
Indication of Success
We have so far completed the following activities:

Architecture Definition.display-selectionquery

display-selectionquerySimilarity Engine.Web InterfaceEl Nino

mediator
Project Impact
The activities outlined above constituted the topic of a Ph.D. dissertation of a graduate student at UCSD. Portions of the same activities provided material for class projects for both undergraduate and graduate students.
A patent application has been requested for the interface design.
The accomplishment of this project spawned other activities related to image database. We have an ongoing project on Virtual Museums which relies heavily on the technology developed for this activity.
Project References
Simone Santini, Explorations in Image Databases. Ph.D. Thesis, University of California, San Diego, 1998
Amarnath Gupta, Simone Santini, and Ramesh Jain, "In Search of Information in Visual Media," Communication of the ACM, 40(12):35--42, December 1997
Simone Santini and Ramesh Jain, "Similarity is a Geometer," Multimedia Tools and ApplicationsSimone Santini and Ramesh Jain, "Image Databases are not Databases with Images," Proceedings of the 9th International Conference on Image Analysis and Processing,Simone Santini and Ramesh Jain, "Do Images Mean Anything?," Proceedings of the 1997 International Conference on Image Processing, October 1997, Santa Barbara, CA
Simone Santini and Ramesh Jain, "Visual Navigation in Perceptual Databases," Proceedings of Visual 97; International Conference on Visual Information Systems, Area Background Databases access information is symbolic form. A database record is a perfectly specified piece of information with a semantics that is completely determined by the syntax of the record.
Visual information, on the other hand, is unstructured information whose semantics is not uniquely specified by the pixel data. In the case of an image, contextImage databases help the user accessing information in visual form. Due to the semantic uncertainty and context-dependency we just mentioned, the process of searching for information in visual data is very different from the process of searching information in traditional databases. The field of image databases tries to idensify and solve the problems posed by visual information search. These problems are not limited to accessing semantics. Images are large two-dimensional objects, and searching images implies dealing with problems like compression, feature extraction, pattern recognition, indexing.

Area References

Myron Flickner Harpreet Sawhney Wayne Niblack and Jonathan Ashley Qian Huang Byron Dom Monika Gorkani Jim Hafner Denis Lee Dragutin Petkovic David Steele and Peter Yanker, Query by Image and Video content: the QBIC system IEEE Computer, 1995
Bach, J.R., Paul, S. and Jain, R, A visual information management system for the interactive retrieval of faces IEEE Transactions on Knowledge and Data Engineering, 5(4) 1993,
Jacopo Corridoni, Alberto Del Bimbo, SIlvio De Magistris, Querying and retrieving pictorial data using semantics induced by colour quality and arrangement, Proceedings of the Third IEEE International Conference on Multimedia Computing and Systems, Hiroshima, Japan 1996