FAST LEARNING AUTOMATON-BASED IMAGE EXAMINATION AND RETRIEVAL

In this paper we study the Image Examination and Retrieval Problem (IE RP). Consider the scenario in which a user wants to browse through a d atabase of images so as to retrieve a particular image which he/she is interested in. Rather than specifying the target image textually, we instead permit the user to access the image by using his/her subjectiv e discrimination of how it resembles other images that are presented b y the system. The IERP is not merely viewed as one involving recogniti on or classification, but instead as one that falls in the domain of c lassifying and partitioning the set of images in terms of their 'visua l' resemblances. In the process, we intend to not merely find images t hat match other images, but, in fact, to group all similar images toge ther so that subsequent searches will be enhanced. The intelligent par titioning of the image database is done adaptively on the basis of the statistical properties of the user's query patterns. This is achieved using learning automata and does not involve the evaluation of any st atistics.