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

One Dimensional Automatic Text/ Picture Image Recognition Algorithm

IP.com Disclosure Number: IPCOM000048738D
Original Publication Date: 1982-Mar-01
Included in the Prior Art Database: 2005-Feb-09
Document File: 1 page(s) / 13K

Publishing Venue

IBM

Related People

Fox, SJ: AUTHOR [+2]

Abstract

A two-dimensional approach in processing mixed format documents having both picture and text information gives accurate results. However, the two-dimensional approach requires additional cost in multiline random access memory (RAM) storage. This article describes a technique whereby many of the advantages of the two-dimensional approach are included with only a minor increment in cost to a one-dimensional approach.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 1

One Dimensional Automatic Text/ Picture Image Recognition Algorithm

A two-dimensional approach in processing mixed format documents having both picture and text information gives accurate results. However, the two- dimensional approach requires additional cost in multiline random access memory (RAM) storage. This article describes a technique whereby many of the advantages of the two-dimensional approach are included with only a minor increment in cost to a one-dimensional approach.

The technique described here is based on the observation that pictures generally have a much greater spacial extent than text. In the majority of documents, only a limited number of pictures (e.g. 1 to 5) can fit across the width of a page. Therefore, this technique requires that in processing a given scan line, the position of any pictorial matter which is detected be recorded by storing the starting point coordinate and an associated run length code. Whenever the one dimensional algorithm detects a transition from text to picture and then back from picture to text, a run length code is stored. The code would be representative of the width of the picture. Unless the code is overridden by subsequent data, it is retained for use in thresholding succeeding lines. This process requires only a minimum amount of storage. If a run length code is less than a predetermined minimum length, it is supressed and the information is processed as text. In addition, alternating the scan direction between left-to-right and right-to-left will also improve the accuracy of the run length codes.

Typical examples of situations where this technique is of value are as follows: Long Horizontal Lines In processing mixed picture/text documents, i...