Optical filter for character recognition - with pattern based on weighted intensities of Fourier transformations to improve resolution

摘要

An optical filter, for use in a character recognition system using Fourier transforms of the images of the objects, is based on the selection of two groups of possible patterns. The distribution of the transparent and opaque sections of the filter is determined by being transparent at points where the ratio of the sums of the weighted intensities of the Fourier transforms of the first group to the sum of the weighted intensities of the transforms of the second group is greater than unity. The result is to improve the resolution between possible patterns with similar transforms, and to simplify the equipment overall.