Browse Prior Art Database

Fast Technique for Finding the Amount Field

IP.com Disclosure Number: IPCOM000110802D
Original Publication Date: 1994-Jan-01
Included in the Prior Art Database: 2005-Mar-26
Document File: 2 page(s) / 83K

Publishing Venue

IBM

Related People

Norton, HT: AUTHOR

Abstract

A fast technique for finding the amount field within personal check images involves the following four steps:

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

Fast Technique for Finding the Amount Field

      A fast technique for finding the amount field within personal
check images involves the following four steps:

1.  finding the most probable horizontal band,

2.  finding the likely dollar sign within that horizontal band,

3.  delimiting the amount field subsequent to that dollar sign, and

4.  recursing through the above steps, as necessary, if the correct
    field is not found in the first iteration.

The process involved in each of the first three steps is as follows:

Step 1 - Find the Most Probable Horizontal Band

The total number of black pels in a sampling of all rows within the
likely region of a check is determined.  These totals are examined to
determine horizontal bands containing non-white portions of the
image.  The horizontal bands are further examined to discard bands
that are too narrow (probably contain pre-printing) and to select
that band that most probably contains the amount field.  If the
selected band is later found to not contain the amount field, the
next most likely band can be selected for further processing.

The examination of only a sampling of rows within the image is
critical to the speed of the process and is adequate to locate the
horizontal band.  In one implementation, a sampling of one row in
five using an image density of 240 pels per inch was found to be
adequate to locate the appropriate horizontal band.

Step 2 - Finding the Dollar Sign Within the Horizontal Band

The image within the selected horizontal band is summed vertically to
find the regions containing non-white portions of the image within
the horizontal band.  The dollar sign usually exists as a reasonably
isolated character within the horizontal band, within experientially
determined bounds of width and density.  The character within the
horizontal band that most closely matches the expected
characteristics is chosen as the probable dollar sign.  If the
selected character is subsequently determined to not be the dollar
sign, the next most likely character location within the horizontal
band can be selected for further processing.

Step 3 - Delimiting the Amount Field Following the Dollar Sign

A more refined examination of the area following the dollar sign, in
both a vertical...