Browse Prior Art Database

Recognition Assisted Segmentation

IP.com Disclosure Number: IPCOM000052057D
Original Publication Date: 1981-Apr-01
Included in the Prior Art Database: 2005-Feb-11
Document File: 2 page(s) / 59K

Publishing Venue

IBM

Related People

Jih, CR: AUTHOR

Abstract

This article relates to a method for determining the segmentation regio in the character string of proportionally spaced type fonts. The method steps comprise: (1) locating the left-hand character edge; (2) executing a predetermined number of vertical scans in the region of the character contiguous to the left-hand edge; (3) applying the scanned subarrays of information to a recognition logic; (4) ascertaining the class to which the character is a member from the scanned subarray information; and (5) determining the right-hand edge of the character based upon the characteristics of the class to which it belongs.

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 55% of the total text.

Page 1 of 2

Recognition Assisted Segmentation

This article relates to a method for determining the segmentation regio in the character string of proportionally spaced type fonts. The method steps comprise:
(1) locating the left-hand character edge; (2) executing a predetermined number of vertical scans in the region of the character contiguous to the left-hand edge;
(3) applying the scanned subarrays of information to a recognition logic; (4) ascertaining the class to which the character is a member from the scanned subarray information; and (5) determining the right-hand edge of the character based upon the characteristics of the class to which it belongs.

In the past, segmentation of characters in a stream required locating both the leading and trailing edges thereof. However, where characters can be partitioned into classes, then the trailing edge will be uniquely determined from knowledge of the leading edge and character class itself. This involves partial image recognition being used before normal segmentation. After the character identity or classes predicted, a segmentation region can be set up according to the normal average size of the characters. Reference can be made to Fig. 1 for a depiction of the functional flow of the method steps.

Fig. 2 shows the hardware implementation of the proposed recognition-assisted segmentation for a character recognition system. As the th of image is generated, each pel (picture element) is temporarily stored in a buffer which is o...