INTRODUCTION
From
time to time we all run across treasured magazine articles, many long out
of print, that have great historical significance to a community of fellow
enthusiasts, that we would like
to share with others online. Unfortunately,
scans of entire 8"x10” magazine pages, with images included, turn
out to be too large in file size to publish on the internet or even send via
e-mail as attachments.
The
good news is that by using OCR (optical character recognition) technology,
the huge file size of scanned text can be reduced down to a mere fraction
in size by converting the scanned image of the text page into editable text
such as a word processor can open and print. The following article offers
practical advice to those who wish to convert book, magazine and newspaper
text articles into editable text conversions and compact graphics suitable
for publication on the web or printer output.
When
scanning any printed material for OCR processing, with mixed text, photo and/or
illustrations, for eventual republication in HTML format, 2 separate scanning
stages must be performed for optimum results. The aim of this article is to
explain the basics of these 2 separate steps.
Step
One: All text pages to be OCR'ed must be scanned and saved using different
scanner control settings than what will be used for the photos and continuous
tone graphics on the same pages.
Second,
all photos and continuous tone graphics in a paper based publication must
be scanned using a different set of scanner settings, plus descreening features
(when available), in order for them to be reproduced small enough for a browser
to view on the web or to be sent as e-mail attachments over relatively slow
modem connections.
Scanning Guideline Basics for all scans:
-
Place top or bottom edge of magazine article
aligned as square as possible to the top edge of the flatbed scanner’s
glass area. If page is cocked
more than 3% out of square alignment, text may not OCR accurately and
graphics may become degraded if later rotated to corrected skewed angles.
-
If available, it is wise to place a black sheet
of art paper behind the text and photo page being scanned to avoid bleed
through of text and images from the opposite side of thin magazine paper, If this is not
done, a ghost like bleed through image of text and graphics on the backside
of the page will most likely appear in the scanned image, thus degrading
the usefulness of that image for both OCR and visual sharing with others.
-
To reduce the scanner image's final output
file size, the scanner's cropping tool can be used to crop out excess
page margins or non-article related sections such as advertisements on
the page.
-
Do not
digitally "sharpen" image during scanning or afterwards if scanning for OCR purposes or
if you intend to edit an image in a image editing program such as Photoshop.
Sharpening can permanently corrupt the image file for use afterwards.
Sharpening should only be as the last step before publishing for online
viewing.
Scanning body text sections for OCR:
All
areas of the page that have blocks of body text (not headlines or fancy type),
tables of data and black & white line illustrations, need to be scanned
at the following settings in order for the OCR program to open and convert
them to editable text.
Scanner
control software settings for OCR conversions:
-
Scale= 100%
-
Scan Mode= line art, b&w or greyscale mode.
Usually the line art or b&w modes are best for e-mail attachments
because the resultant image file for an 8x10 page will be smallest.
-
Scan resolution=300 dpi
-
Descreen or anti-moire settings= off
- File format to save in= .TIF , no
LZW compression applied. Saving
in any other format or with LZW compression, may render image file useless
for OCR purposes.
Scanning Photos for Web Display:
Because
graphics for the web do not display at any higher resolution than 72 dpi on
a common computer monitor, color, black and white photos and continuous tone
illustrations do not need to be scanned at any resolution finer than 72 dpi. However, some web designers and graphic
prep specialists prefer to scan at twice that resolution (150 dpi) and then
scale down to 72 dpi just before saving them for final display web graphics
because they feel that the final image will look cleaner.
Due
to 56k modem bandwidth limitations of sending such scans quickly over the
internet as e-mail attachments, most modem users would be wise to scan at
72 dpi and save in either .gif or .jpg format to keep file size to a minimum. If both sender and receiver have high
bandwidth internet connections with adequate mailbox sizes, then scanning
at 72-150 dpi and saving the scans int., .eps or .psd graphic formats will
result in a superior quality web display graphic.
Scanner control
software settings for Photos & continuous tone graphics
-
Scale= 100%
-
Scan Mode= Color or B&W Photo or Continuous
Tone mode.
-
Scan resolution=72 dpi for web display (150
dpi for high bandwidth attachments, 300+ dpi for printer output)
-
Descreen or anti-moire settings= ON Your scanner software
may use other terms for this technical feature. Read the scanning FAQ for more information about de-screening.
- File format to save in= .GIF for
line art, .JPG for photos and continuous tone images. Save at Photoshop level 8-10 or “high
quality” settings (for non-web use & printer output save as .eps,
.tif or .psd format)
CAUTION:
DO NOT DIGITALLY "SHARPEN" SCANNED IMAGES IF INTENDED FOR OCR
PROCESSING OR FURTHER DIGITAL EDITING. YOU COULD CORRUPT THE IMAGE PERMANENTLY,
YOU COULD RENDER IT NON-EDITABLE AFTERWARDS. ONLY SHARPEN
AN IMAGE JUST BEFORE YOU INTEND TO PUBLISH IT ONLINE. WHEN SENDING YOUR IMAGES
TO A WEB DESIGNER OR PUBLISHER, NEVER SHARPEN YOUR IMAGES. LET THEM DO THAT
WHEN THEY THINK BEST.
Scanning and article conversion FAQ.
What is descreening? Unlike photographs
printed by your local photo finisher, photos and continuous tone images printed
by a printing press such those in newspapers, magazines and books, are actually
made of thousands of very small ink dots referred to as "halftone dots".
Because a digital image, such as what a scanner generates, is also made of
small dots of it's own, a strange visual phenomenon can result to create a
juxtaposition of these 2 dot patterns creating an optical illusion effect
called moiré. Descreen
technology attempts to digitally rearrange the relative proximity of these
2 dot patterns in such a way as to cause the moiré effect to disappear
or decrease from the final scanned image.
Halftone dot and moiré elimination features
included in the scanner software control programs bundled with most popular
scanners are well worth investigating in order to avoid ugly moiré
and hatch patterns being overlaid on top of attractive graphics as a result
of scanning without descreen technology applied.
When
a scanner's desreening features work well and the moiré pattern overlay
is removed, the visual quality of the final scan will appear to match that
of the original printed image. If your first attempt to use your scanner's
descreen feature does not correct the problem sufficiently, try rotating the
paper image slightly around an imaginary center axis on the scanner's glass
surface and scan again. Sometimes
the the descreen feature will be more effective after these slight rotations.
Descreen
features that come bundled with quality scanners often have multiple configuration
settings. Reviewing your scanner
software manual may reveal that you have these very powerful extra features
that may provide you with more descreening control than you first suspected. Be sure to read your manual.
Alternatives
to De-screening
My
friends, Tom and Greg, have also reminded me of an alternative method that
I had forgotten to mention. By scanning at higher resolutions, usually around
300 dpi, and then reducing the image file's resolution down to 72 dpi using
an image editing program, such as Photoshop or Paint Shop Pro, a person can
often create scans of half-toned images that will display with little or no
moire or hatched patterns visible across the 72 dpi image afterwards. Some
experimentation may be required for best results, but it can be worth the
effort.
Another
alternative is to use special filters found in some image editing programs
that attempt to detect and correct moire and hatch patterns in a scanned image.
I have found the results to be somewhat hit and miss.
How much to scan at one time?
-
TEXT:
Scan one page at a time in page order, later naming each scan by article
name abbreviation plus article page number, i.e. HR240Zp1, HR240Zp2, etc.
Usually, for consistent quality OCR conversions, it is best to scan
the entire text area on one page including any photos and illustrations
embedded within the text or text columns. Later, the graphics can
be deleted from the text page with an image editor before the OCR program
attempts to “recognize” and convert the text on that page.
If
all text pages are cropped, scanned and named in page order, the OCR program
will be more likely to rebuild the formatting of the text article accurately.
After the text is successfully converted, a program such, as a word processor
or web authoring program, can insert the article separately scanned graphics
into their original positions.
-
PHOTOS
& GRAPHICS: If the
resultant scanned image files are not too large, entire blocks of photos
on one page can be scanned as a group ( example
) and later cut up in an image editing program, thus making it quicker
for the person scanning the article to complete his task. To reduce overall image file sizes and to make it easier to
transmit as e-mail attachments, the scan operator can also use the scanner’s
selective cropping tool to scan and name each graphic separately. It is best to scan and name them in the
order that they appear in the article so that a web designer can reassemble
them into an web page layout in proper relationship to the text. If the graphics on the page are labeled,
scan the label as part of the scan of the graphic for use as positioning
reference later on.
-
Photos
and illustrations that span two pages should be scanned as one image
if pages can be made to lay perfectly flat and in parallel alignment.
If article pages are loose, it is wise to tape them together from
the back side using easy to remove, removable Scotch Tape.
Don’t worry about the crease line or break that runs down
between the 2 sides of the image.
This area can be reproduced for realism or can be rubber stamped
in Photoshop to blend it into the look of the image.
Scanning article photos
for digital printing instead of web display. For best results scan
all graphics for printer output at 300 dpi or higher or images may reproduce
as jagged or fuzzy looking rather than sharp. For best printing quality, scanned image files should be saved
in .tif, eps or .psd format rather than .gif or .jpg.
©2001
Eric Neyerlin - all rights reserved
.
No
part of this article may be displayed or reproduced in any form of written,
digital or online media nor sold in any form without written permission of
the author. This restriction also applies to any website attempting to display
this page within a frame of another website not authorized by the author to
do so.
Last
Updated January 30, 2004
