Component-based architectures for computer vision systems

Abstract

Research performed in the field of computer vision has steadily ignored recent advances in programming tools and techniques, relying on well-established traditional methods, such as Unix-based C programming. While this can certainly be effective, modern computer vision research may benefit significantly from the new tools and technologies that have recently become available. This paper addresses the use of component-based programming methods and proposes a model loosely based on 3-tier architectures, for the creation of robust and reusable computer vision systems, in order to improve code modularity and reusability, and to ultimately foster cooperation between researchers in the field. It outlines a basic design strategy and exposes the benefits and drawbacks of migrating to component-based code. The model is used to build a component-driven framework that is designed based on the principles of 3-tier applications. Its purpose is to aid in the creation and maintenance of stable, dependable testing and development environments. We have listed the main advantages of this approach and have concluded that although the learning curve for the programming skills required is steep, the benefits to be reaped are worth it.