Parallelization of an image analysis application: problems, results and a solution framework

The implementation of image analysis applications consisting of many tasks on parallel machines raises problems regarding compatibility of data-representations between tasks and allocation of available resources. These problems arise from the integration of different parallel algorithms in one system and are not raised when parallel algorithms are developed for each task independently. The harmonization of data-representation requirements of different parallel algorithms and the partitioning of resources among them are problems that must be addressed at the level of application and not at the level of individual tasks. In this paper the problems are formulated, components of a system-based approach to these problems are presented and solutions are suggested that facilitate the development of complex industrial image analysis applications on parallel machines. The concrete example is a text-recognition application