Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Memory Reduction Method for Multi-Template Optical Character Recognition System with Variance Data

IP.com Disclosure Number: IPCOM000114169D
Original Publication Date: 1994-Nov-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 2 page(s) / 36K

Publishing Venue

IBM

Related People

Mano, T: AUTHOR [+2]

Abstract

Disclosed is a recognition library size reduction method for multi-template Optical Character Recognition (OCR) systems with variance data merging variance data for a same character. Generally speaking, the recognition accuracy of a multi-template OCR system is significantly improved by using variance data of each template. On the other hand, the amount size of variance data is big and it is memory consuming. This paper describes a method for the memory reduction without degrading the recognition accuracy.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 78% of the total text.

Memory Reduction Method for Multi-Template Optical Character Recognition
System with Variance Data

      Disclosed is a recognition library size reduction method for
multi-template Optical Character Recognition (OCR) systems with
variance data merging variance data for a same character.  Generally
speaking, the recognition accuracy of a multi-template OCR system is
significantly improved by using variance data of each template.  On
the other hand, the amount size of variance data is big and it is
memory consuming.  This paper describes a method for the memory
reduction without degrading the recognition accuracy.

      In conventional template matching methods with multi-templates
with variance data for each template, the total library size is
          N * F * M               template data size
          N * F * M               variance data size
   -----------------------
          N * F * M * 2           total library size
where N is the number of supported characters, F is the size of
feature data, and M is the average number of multi templates.  To
reduce the library size, variance data for a same character are
merged.  (For example, the merge process is done by taking average
values.)  Then a new library size is as follows:
          N * F * M               template data size
          N * F * 1               variance data size
   ----------...