An Architecture for Asynchronous Indexing of Media Assets

Multimedia documents, though rich in content, generally lack structured and descriptive metadata that would allow indexing, searching, and cross-linking. Since the field of Multimedia Information Retrieval has been working on solving the problem of multimedia content analyzing and indexing for a number of years, several existing components could be integrated into a final solution, as well as incorporating new techniques from current research results. Hence, a multimedia indexing system must not only face the challenge of a providing scalability and performance, but should also allow for interoperability with legacy code and integration of new processes. With these goals in mind, this paper presents a loosely coupled, asynchronous, distributed event-based SOA media indexing system, making use of Web services and Enterprise Service Bus technologies and contrasts it with different application distribution approaches, with regards to the throughput, scalability, and response time when executing the tasks required for analyzing and indexing of large media document collections.