Mining Image Datasets using Perceptual Association Rules

J. Tesic, Shawn Newsam, and B.S. Manjunath

Dept. of Electrical and Computer Engineering
University of California at Santa Barbara
Santa Barbara, CA 93106
Email: {jelena, snewsam, manj} [at]


This paper describes a framework for applying traditional data mining techniques to the non-traditional domain of image datasets for the purpose of knowledge discovery. In particular, perceptual association rules, a novel extension of traditional association rules, are used to distill the frequent perceptual events in large image datasets in order to dis- cover interesting patterns. The focus is on spatial associa- tions although the method is equally applicable to associa- tions within or between other dimensions; i.e., spectral, or in the case of video, temporal. A primary contribution is the derivation of image equivalents for the traditional associa- tion rule components, namely the items, the itemsets, and the rules. The proposed approach is modular, consisting of three steps that can be individually adapted to a particular application. First, the image dataset is labeled in a per- ceptually meaningful way using a visual thesaurus. Second, the firrst- and second-order associations are tabulated in a scalable data structure termed a spatial event cube. Finally, the higher-order associations and rules are determined using an adaptation of the Apriori algorithm. The proposed ap- proach is applied to an aerial video dataset to demonstrate the kinds of knowledge perceptual association rules can help discover.
[PDF] [BibTex]
J. Tešic, S. Newsam and B. S. Manjunath,
SIAM Sixth Workshop on Mining Scientific and Engineering Datasets in conjunction with the Third SIAM International Conference (SDM), San Francisco, California, May. 2003.
Node ID: 359 , DB ID: 157 , VRLID: 114 , Lab: VRL , Target: Proceedings
Subject: [Multimedia Database Mining] « Look up more