Designing Image Processing Algorithms on Cell Broadband Engine: Programming strategies and performance analysis

Abstract

The multicore processors, which are widely available in recent times, have great potential to achieve significant performance improvement of the Digital Image Processing Algorithms. Due to the availability of inherent parallelism, large number of resources and functionality on different cores of such multicore architecture, significant performance gain can be obtained by implementing applications on it. But the major challenge lies in programming and designing applications on such multicore architecture to exploit the available resources and functionality to its fullest. These challenges are addressed by developing two image processing algorithms with different computational load on Cell Broadband Engine - a hybrid multicore processor from Sony, IBM and Toshiba. The performance analysis of different versions of two algorithms is shown. The programming approach, designing strategies and drawbacks in development of image processing algorithm on Cell Broadband Engine are addressed.