Towards an efficient high-level modeling of heterogeneous image processing systems

High-level formal modeling is a beneficial method for assisting system architects in early performance evaluations of desired hardware/software architectures and for coping with the continuously increasing complexity of heterogeneous systems. Modeling languages like UML allow expressive and concise specifications of algorithms, platforms, and infrastructure artifacts for different domains. Techniques are already available for deriving implementations but also executable simulations from UML-based specification models. We use such derived simulations for early investigation and validation of the behavior of the designed system. However, building specification models is often a very complex and time-consuming process, which highly depends on the designer's knowledge, used development process and modeling languages. In this paper, we introduce a holistic top-down modeling methodology developed for enabling an efficient design of heterogeneous image processing architectures using the standardized UML modeling language. Particularly, the suggested methodology focuses on improving user guidance with clear modeling workflows to achieve a less error-prone model design process and to reduce the system development effort. Moreover, we demonstrate how the specific domain knowledge can be captured directly in a UML conform modeling library and that this library enables us to express image processing algorithms in an efficient manner.