Efficient Access based on the Feature Origin


Jelena Tesic, Sitaram Bhagavathy, B.S.Manjunath


In typical applications, image/video descriptors are of high dimensionality (from tens to several hundreds). Color, texture, shape and motion descriptors are some of the commonly used low-level visual descriptors for image and video data. For example, a 256-dimensional color histogram descriptor is often used to characterize the color distribution in a given image. The feature space grows exponentially with the dimensions, and the search complexity increases at the same rate. In high feature dimensions, the curse of dimensionality is an issue as the traditional database indexing methods and clustering methods do not scale well beyond 10--20 dimensions. This poses challenging problems for database access. Therefore, the high dimensionality and computational complexity of this descriptor adversely affect the efficiency of content-based retrieval systems.

We propose a modified MPEG7 texture descriptor that has comparable performance, but with nearly half the dimensionality and less computational expense. Furthermore, it is easy to compute the new feature using the old one, without having to repeat the computationally expensive filtering step. We also propose a new normalization adaptive indexing methods that improve similarity retrieval and a bit allocation compression scheme for indexing that improves search efficiency up to 10 times. New descriptor and indexing structure are evaluated over range of scientific datasets.


These materials are presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each authors copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder. Jelena Tesic, Sitaram Bhagavathy, and B.S. Manjunath, "Efficient Access based on Feature Origin," to be submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence, November 2003. J. Tesic, S. Bhagavathy, and B. S. Manjunath, "Issues Concerning Dimensionality and Similarity Search," Proceedings of 3rd International Symposium on Image and Signal Processing and Analysis (ISPA), September 2003, Rome, Italy, September 2003. [abstract] S. Bhagavathy, J. Tesic, and B. S. Manjunath, "On the Rayleigh nature of Gabor filter outputs," IEEE International Conference on Image Processing (ICIP), Barcelona, Spain, September 2003. [abstract]


This research was supported in part by The Institute of Scientific Computing Research (ISCR) award under the auspices of the U.S. Department of Energy by the Lawrence Livermore National Laboratory under contract No.W-7405- ENG-48, ONR# N00014-01-1-0391, NSF Instrumentation #EIA-9986057, and NSF Infrastructure #EIA-0080134.