A data-flow multiprocessor architecture for the transform coding of regularly decomposed images

✍ Scribed by N.A. Alexandridis; P.D. Tsanakas; S.V. Konstantinidou

• Publisher: Elsevier Science
• Year: 1985
• Tongue: English
• Weight: 839 KB
• Volume: 27
• Category: Article
• ISSN: 0378-4754
• DOI: 10.1016/0378-4754(85)90078-3

✦ Synopsis

An architecture for the parallel implementation of an orthogonal transform of images is presented. Images are preprocessed by the "regular decomposition" procedure and converted into a quad-tree representation. The orthogonal transformation is performed progressivelly by adjusting to the various hierarchies of the quad-tree nodes. The architecture of the proposed machine is based on the data-driven model of computation in which an operation iS executed when its operands are available. The specific requirements of the image processing algorithms have been taken into account, principally the efficient handling of array data--structures. The tagged-token method supports recursive operations and helps to better exploit the inherent parallelism of the transformation algorithms.