METHOD FOR IDENTIFICATION AND SEGMENTATION OF TOUCHING CHARACTERS

摘要

Characters which touch on a written document are identified and separated according to their contours. The elements appearing in the document are divided into two categories respectively including those with internal contours closed and those that do not possess. A member of the first category having either a closed inner contour that is not centered with respect to the element, two internal closed contours horizontally offset, or more than two internal closed contour is identified as containing two characters which touch and is divided according to the thickness of the character bars observed in the document. A member of the second category whose contours intersect three horizontal reference lines spaced in a number of points greater than a predetermined number is compared with the intersection models characteristics of individual characters. If no match is found, the element is considered to have characters that touch.