Textural segmentation using an optical scanning technique

Abstract

Two types of filter capable of extracting the boundary between textural domains modulated by different spatial frequencies have recently been proposed.1,2 These filters are extrapolations of tonal edge detection filters: the odd-type (w-Gaussian) and the even-type (w2-Gausslan). We first review briefly the model of a textural edge In an image and the textural edge extraction filters. We then describe an optical scanning system capable of making both filters. It Is shown that the textural edge extraction filter based on the odd-type tonal edge detection filter produces an output with one single peak, whereas that based on the even-type filter produces two peaks as an output.