Abstract
One of the biggest problems facing optical signal processing is the low accuracy of most optical processors. Threshold-decomposition image processing overcomes this problem to a large degree in the performance of nonlinear filtering operations, such as morphological and ranked-order filtering, on gray-scale images. Threshold decomposition is a technique by which an input gray-scale image is decomposed into a series of binary cross sections, each corresponding to the input thresholded at a different level. The binary cross sections are individually convolved with a binary kernel and are subjected to a thresholding/hard-limiting operation (the key to the noise immunity), and the results are summed to produce a filtered gray-scale image. The robustness of this process to system noise implies that fast, parallel optical processing technology can be combined with digital electronics without degrading the overall system accuracy. The implementation of such an opto-electronic system would consist of an emitter (e.g., laser diode) array, a photodetector array, an incoherent optical convolver, and digital electronics. We analyze the immunity to noise generated by spatial variations of emitter and photodetector arrays in the performance of the morphological operations of median filtering, dilation, and erosion.
© 1992 Optical Society of America
PDF ArticleMore Like This
Tien-Hsin Chao
ThBB6 OSA Annual Meeting (FIO) 1992
James M. Hereford and William T. Rhodes
THM5 OSA Annual Meeting (FIO) 1989
Ravindra A. Athale, Joseph N. Mait, and Dennis W. Prather
TuB3 Optical Computing (IP) 1991