Parallel programming for real-time image processing using computing agents

Scalable, maintainable, and reusable algorithm development and flexible dynamic computing resource allocation are key issues in using Multiple Instruction Multiple Data (MIMD) computing platforms for large-scale real-time embedded applications.. A novel parallel programming paradigm for such systems, agent-oriented parallel programming, is proposed in this paper. The proposed methodology attempts to address four major issues on the application of MIMD parallel computing platforms: scalability, maintainability, reusability, and fast dynamic resource allocation. Application of the proposed method on a real-time image processing system shows that the method is very effective. Although the method is proposed mainly for embedded applications, the concepts can be used for other distributed computing systems.