Distributed Implementation of Discrete Cosine Transform Algorithm on a Network of Workstations

Abstract. The paper presents a distributed platform based on CORBA middleware that allows obtaining high performances in execution of Discrete Cosine Transform (DCT) algorithm. Ideally suited for image compression, DCT is a very intensive computational task, which can be cost-effective solved in a network of workstations (NOW). The transform algorithm is partitioned into a number of independent tasks that are assigned to a group of components (servers), distributed on the hosts of a NOW. Adjusting the number of image blocks assigned to each server assures efficient communication of data through the network and computation load balancing. The experimental results show significant speedup in distributed execution of this application and, also, that the performances are improved with increasing data size, making distributed algorithm particularly suitable for large images