Geometric transforms on parallel architecture
This paper shows that geometric transforms may be performed on a Line Processor with performances satisfying actual real-time constraints. In order to parallelize such transforms on the Line Processor SYMPATI 2 we take advantage of its access capabilities for processing the image either in a row by row or in a column by column fashion. These capabilities allow us to decompose image transforms into 2 steps. The first one performs a x-constant transform, the second one a y-constant transform. This decomposition makes it possible to obtain very attractive processing times in comparison with those provided by wired operators.
- JUVIN D., BASILLE J.-L., ESSAFI H., LATIL J.-Y: “SYMPATI 2, a 1,5 D Processor array for image application. “Eusipco Signal Processing IV: Theories and applications, North-Holland, pp 311–314, 1988.Google Scholar
- BASILLE J.-L.: “Traitement d'images numériques, application: système inter-actif de caryotypie”, Thèse de Docteur Ingénieur, Université Paul Sabatier, Toulouse, 1976.Google Scholar
- C. F. R. WEIMAN: “Highly parallel Digitized Geometric Transformations Without Matrix Multiplication”, International Joint Conference on Parallel Processing, pp. 1–10, 1976.Google Scholar
- C. F. R. WEIMAN: “Continuous Anti-aliased Rotation and Zoom of Raster Images”, Computer Graphics Vol. 14(3), pp. 286–293, Juillet 1980.Google Scholar
- K. A. CLARKE and H. H-S. Ip: “A Parallel Implementation of Geometric Transformations”, Image Processing Group, Department of Physics and Astronomy, University College London, Report No. 82/5, March 1982.Google Scholar