A parallel processing framework for real-time software-only video compression for broadcast and mobile services

Abstract

This paper develops a generic parallel processing framework on embedded multicomputer systems for real-time video compression for convergent broadcast and cellular services. This framework allows easy prototyping of new parallel video compression algorithms and integration with video applications. Using the proposed parallel processing framework, a parallel H.263 encoder has been implemented on a PowerPC processor based Mercury RACE multicomputer. The experimental results show very promising parallel speedups. Real-time performance was achieved for QCIF video using the non-optimised reference software. The parallel processing framework not only can be configured with different number of processors on an embedded multicomputer system but it is also portable onto varied embedded multicomputer platforms. Therefore, this paper provides a software implementation solution as an alternative to hardware video compression for broadcast and mobile services.