Efficient Image Retrieval on Distributed Cluster Architectures

Abstract

The Internet technology allows access to a huge amount of digital information, represented by text, images, graphics, audio and video sequences. The Internet can thus be considered a genera] digital library, containing a collection of comprehensive knowledge, which is distributed over millions of independent nodes. This creates an urgent need to organise, manage, and to retrieve multimedia objects. However, the large memory, bandwidth, and computational requirements of the multimedia data surpass the capabilities of traditional database systems and architectures. The performance bottlenecks can be avoided by partitioning the data among multiple nodes and thus creating a configuration supporting parallel storage and processing. A reference system model, methods for media distribution, strategies for workload balancing and for data backup are presented in the article.
The methods introduced here are useable for all kinds of multimedia objects. We will focus on image retrieval and image databases, which are an important sub-system of multimedia information systems, in order to provide a clear and compact presentation of the necessary methods for multimedia content analysis, description, and retrieval.