An effective parallelizing scheme of MPEG-1 video encoding on Ethernet-connected workstations

Abstract

Although MPEG-1 Video is a promising and the most widely used moving picture compression standard it requires a lot of computational resources to encode the moving pictures with a reasonable frame size and quality. In this paper we propose and implement an efficient parallelizing scheme for an MPEG-1 Video encoding algorithm on Ethernet-connected workstations which is the most widely available computing environment nowadays. In this parallelizing scheme, the slice-level, frame-level, and GOP (Group of Pictures)-level parallelisms are identified as the attractive parallelisms that can be exploited in Ethernet-connected workstations. Three efficient parallel implementation schemes considering the communication characteristics of Ethernet-connected workstations are also proposed and experimented A series of experiments using thirty workstations shows that the MPEG-1 Video encoding time can be reduced in proportional to the number of workstations used in encoding computations although there is a saturation point in the speedup graphs.