Content-Based Web Image Retrieval via Mobile Multi-Agent Architecture

A novel scalable architecture for automating and speeding up content-based image retrieval (CBIR) on the web is presented by using mobile multi-agent. Traditional web CBIR functions are encapsulated in six agent classes via object-oriented idea. The retrieval domain is modelled as a multi-agent system, where, agent objects operate autonomously while cooperate with each other, and its coordination with outside environment can result in web CBIR. During retrieval, the system can decide an optimized retrieval strategy by automatically analyzing initial query input with its repository, and through feedback, it can also make corresponding adjustment by intelligently deducing user intent and interest, without any human interference if necessary. Furthermore, the mobility enables the search agent to perform local image match at database source, which is expected to improve retrieval speed and reliability by minimizing network traffic and dependence on network status. In addition, the repository management agent with heterogeneous database classifying and indexing ability is also expected to further extend search range and ensure retrieval efficiency. Simulation results validate the effectiveness of this approach by significantly outperforming its traditional counterparts.