Multi-Tempom Image Segmentation Using a Distributed-Memory Parallel Computer

Abstract

A method is described €or remotely sensed image segmentation based on the integration of edge detection and region growing approaches. The method is designed for full exploitation of parallel processing. The initial edge detection and contour following procedures are perfomed on separate image sub-scenes. These procedures then generate separate and independent sub-areas of the image which can then be used for best-merge region-growing without mutual interference. These sub-areas can be handled by separate computer processors. A network of T-800 transputers is being used for this purpose.