METHOD OF AND APPARATUS FOR OBTAINING BINARY IMAGE

A frequency distribution or histogram M (L.sub.j-1, L.sub.j) of differences between gradation levels (L.sub.j-1, L.sub.j) of adjacent pixels (P.sub.j-1, P.sub.j) in an image is obtained, and the number of level-changes for each tentative threshold level L.sub.j is determined as a function of the histogram of the level differencies. By using the number of level-changes, an optimum binarization threshold level is set by means of an adaptive thresholding method. Since the number of level-changes is calculated from the histogram of the level-differencies without practical binarization for each threshold level, the amount of calculations is not dependent on the number of pixels but on the number of gradation levels. The calculations for setting the threshold level for the binarization of the image are achieved at high speeds by a small-scale hardware.