Video normalization for hand print recognition

Abstract

In the machine recognition of hand printed characters it is desirable that the characters be constrained to a uniform size. This may be accomplished by taking a relatively unconstrained hand printed character and, through simple logical processes involving eliminating and/or combining portions of the unconstrained character, obtain a constrained character of a specified size. To obtain this specified character, the unconstrained characters are stored in a video line buffer as a plurality of rows of binary bits. When a row character is located in the video line buffer a process of vertical and horizontal consolidation is initiated. In the vertical consolidation, which is performed first, predetermined rows of the bits of the row character are selectively combined with other rows of stored bits or selected without change to reduce the height of the character to a predetermined desired height. Unselected rows are ignored. As the rows of bits are obtained during vertical consolidation, horizontal consolidation of each row is initiated. This consolidation is performed on a byte by byte basis so that each row is transformed in accordance with a predetermined relationship between the width of the row character and the desired width.