Browse Prior Art Database

Segmentation for Optical Character Recognition

IP.com Disclosure Number: IPCOM000074321D
Original Publication Date: 1971-Apr-01
Included in the Prior Art Database: 2005-Feb-23
Document File: 2 page(s) / 57K

Publishing Venue

IBM

Related People

Pillar, DW: AUTHOR [+2]

Abstract

The separation of serif character fonts is a difficult task for raster-scan character-recognition systems, since a considerable number of the characters overlap in a horizontal direction.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 62% of the total text.

Page 1 of 2

Segmentation for Optical Character Recognition

The separation of serif character fonts is a difficult task for raster-scan character-recognition systems, since a considerable number of the characters overlap in a horizontal direction.

Logic circuit 10 provides a segmentation signal on line 11 when a serif-like "terminating line element" (TLE) is detected. Inputs 12 are derived from the 7th, 8th and 9th scans ahead of the current scan (scan 1) in shift register 13. Inputs 12 are binary "1" or "0", depending upon whether the associated pattern bit is "black" or "white". AND gate 14 and inverters 15 provide a positive output on line 16 when all of the register cells R71, R72, R81, R82, R91 and R92 indicate "white". This output is held in latch 17 until a signal from OR 18 indicates either that cells R91-R93 are all white (AND 19 and inverters 20), or that an end-of- scan signal has appeared on line 21. Latch 22 is set by the confluence in AND 23 of a signal from latch 17, a "black" indication from cell R72 on line 24, a black condition in either of the cells R71 or R73 (OR 25), and a black bit in any of the cells R81-R83 (OR 26). Latch 22 is also reset by OR 18. The unlatched output 16 and output 27 of latch 22 together set latch 28 via AND 29. Segmentation output 11 then provides a signal until latch 28 is reset by a signal on line 21.

When line 11 is active, it may be held for approximately five scans. During this time, it preferably overrides all other less power...