Agent Based Framework for Content Based Image Retrieval

In this paper, we plan to study the viability and effectiveness of using an agent-oriented approach to developing content-based image retrieval (CBIR) systems. Image retrieval is a key technology that significantly aid the work of both professionals (medical practitioners, engineers, etc.) and has the potential of enabling a much richer interaction of the average home user with the distributed information content on the internet and the world-wide-web. CBIR focuses on using feedback from the user to facilitate the search for matching images based on the content of a query image. Some of the most practical use of agent technology results in building personalized agents to assist users with their information processing needs. Such agents have both expertise in the domain of application, e.g., travel planning, financial planning, etc. and a model of the biases, preferences, constraints of the associated user. We believe that an agent-based CBIR system has several advantages over a personal systems in terms of flexibility, ease-of-use, adaptability, transparency, and precision of recall. Our proposed Agent for content Based Image Retrieval (ABIR) is an interactive image retrieval system based on relevance feedback from the user. ABIR will use the preference of the current user, the usage patterns of past users, and novel feature preference combination methods to iteratively interact with the user and guide the search for satisfactory matching images.