The DFrame: Parallel Programming Using a Distributed Framework Implemented in MPI

摘要

High content throughput imaging systems must apply time consuming complex image processing algorithms to multiple bio-medical image streams. These systems are typically designed to use parallel resources in order to achieve results in reasonable time scales. This paper presents the design of a distributed framework that provides separation of the largely orthogonal parallelisation from the domain image processing algorithm development. This allows reuse and pluggable extension of parallelising patterns, as well as providing for extensibility of domain image processing.