Case Study: Cairo - A Distributed Image Retrieval System for Cluster Architectures

Abstract

Cairo is a distributed, cluster-based image retrieval system that provides a high-quality, object-based image analysis and search. The state-of-the-art retrieval approach, which compares entire images, is extended by an exhaustive search in all image sections for the occurrence of selected regions of interest. The large computational effort requires the use of parallel architectures in order to maintain reasonable system response times. The goal of this chapter is to give an overview of the different techniques required for distributed image retrieval, such as querying mechanisms and operations for image analysis and comparison, related aspects of parallel processing, scheduling, and data placement. The proposed architecture, Cairo, combines all these aspects and offers one possible design for a user-friendly, flexible, efficient, and distributed image database. The global system structure, the design of the individual modules, the internal communication, and typical use cases are outlined in the main part of this chapter.