Abstract

Images in the natural sciences often posses distinctive topologies, thus rendering order statistics better suited for image processing than more traditional linear filtering. A useful subclass of order statistics based on binary images is mathematical morphology./1/ Mathematical morphology is also well suited to an optical implementation. /2-5/ Optical mathematical morphology can be performed at a frame rate of 10-100 kHz., thus permitting real-time non-linear image processing in many applications. Our proposed optical architecture also allows for programmable parallel processing of very large images, under control of a small electronic micro-processor.

© 1991 Optical Society of America