Why the OCR is still misreading the text?

Resolving The Problem

QUESTION:

Why is the OCR (Optical Character Recognition) misreading the text?

ANSWER:

Unfortunately, there is nothing more that can be done if the below recommendations are followed.
Enhancements have been entered in our system to improve this functionality in the future releases.
27629 Enhancement - Modify baseline of region OCR verification
26506 Enhancement: Improvement for the OCR feature of the Image Comparator
10167 Better OCR

For information about OCR Tips, open the online Help and search the index for OCR.


From the online help:


OCR Tips

The following tips will help you create OCR regions that test well. For
information on creating OCR regions, see Creating OCR Regions.


· To get the best results from an OCR operation - place the region
rectangle border so that the left side is close to the first character of
the text to be recognized. If there's a lot of space there, it doesn't
operate as well, and sometimes the tool may fail to recognize the text. If
you get an OCR error, use the Zoom commands to zoom in on the area, and move
the left border closer to the first character.

· Fit all the borders of the region as tightly as possible around the
text.

· When you create an OCR region, it is automatically "read." Check the
Mask/OCR List to make sure all the text appears that you want to be tested,
and that no extra text is present. Resize or move the region if necessary.
To eliminate extraneous marks, you may need to make it smaller. To include
all the text you want tested, you may need to make it larger. See Moving and
Resizing OCR Regions.

· If your region contains white or light colored text, use the Light
text option in the OCR Region dialog box. This box automatically opens when
you create an OCR region.

· If your region contains text on a gray or dark colored background,
use the Gray background option in the OCR Region dialog box. This box
automatically opens when you create an OCR region.

· If the OCR region border accidentally includes some part of a
graphic or icon, it may fail to recognize the text within the region. Resize
the region so that you don't include part of a graphic.

· Smaller fonts are harder to recognize. Larger fonts test better.
Also, bold fonts are easier to recognize.

· If you have a colored font on a colored background, play with the
option settings in the OCR Region dialog box to see what works best. The
image is converted to black and white before sending it to be recognized.
You can check the OCR<number>.BMP in the Windows temp directory to see what
was actually produced and to use as a guide in these cases. The number in
this file name corresponds to the number listed in the Mask/OCR List. See
the note below.

· Unless you're depending on and making decisions in script based on
the recognized text, it's not necessarily important that all letters of a
text string are recognized correctly. It's only important that it be done
the same way during recording and playback, so that the results are
accurate. If you notice any characters not being picked up, it may not be
malfunctioning, but may just have a limitation. Chances are it will work
accurately even if every character is not recognized.

· When the OCR engine cannot recognize a character based on its
surroundings, it will insert a '
~' character. Some characters in 8 point MS Sans Serif (default Windows
font) are harder to read than others - those with a blocky shape are easier
than those with angles. This is due to the DPI resolution at which they're
captured versus what is provided to the recognition engine. The characters
appear more jagged at the higher DPI if they have curves.

· We recommend 150 DPI for the best operation of OCR. This is the
default setting found in the Options dialog box.