Abstract

Hybrid optical implementation of a rotating kernel min-max transformation is discussed and applied to optical edge detection and enhancement. An input image is convolved with a short, narrow 2-D kernel that is rotated, either continuously or discretely, through 360°. As the kernel rotates, the convolution output is monitored, and the maximum [= Max(x,y)] and minimum [ = Min(x,y)] values at each point (x,y) are stored. Since the kernel used is typically real and non-negative, the convolution operation can be performed using an incoherent spatial filtering system. Using an asymmetric kernel, two different types of edge images are obtained: (1) a gray-scale edge image and (2) a binary edge image.

© 1990 Optical Society of America