A number of architectural issues raised by hierarchical image processing/transmission algorithms are presented. First, the inherent performance limitations of the conventional parallel architectures are analyzed from the image processing point of view. Then the concepts of the hierarchical image data structures and algorithms are presented along with a brief description of an algorithm for progressive image transmission. It is shown that the most promising source for increasing the systems' throughput is to introduce modularity in the architectures leading to adaptable real-time supersystems that support a dynamic redistribution of the available hardware resources. The proposed alternative aims to more efficient system configurations that are in congruence with the respective hierarchical structures and their reconfiguration requirements for efficient image processing and progressive orthogonal transform coding and transmission.
展开▼