Image Processing Reference
or (R t ,G t ,B t ) = (0,0 + 31,0) or any other combination that would meet the above request, in order
for the human eye to recognize what is writen. Figure 7 proves what is being said before.
FIGURE 7 The condition to be met for the characters to be recognized by the human eye.
A reasonable assumption that is made from the start is that characters that would not be
perceived by the human eye are not expected to be recognized by a machine using OCR, but
every character perceive by the human eye must also be recognized by a machine (assuming
there is no noise), or at least expect to be recognized. Thus, for the text that could not be per-
ceived by the human eye, the performances of OCR are not improved by filtering first the text
image using Highlight ilter.
The core of the algorithm is described in the following. Since an example makes the general
case more explicit, let's bring into discussion the Blue component of the pixel having the co-
ordinates ( x , y ) relative to the upper-left corner of the image.
If the absolute difference between the value of the Blue component of the pixel with co-
ordinates ( x − 1, y − 1) and the value of the Blue component of the pixel having the coordinates
( x , y ) is greater than 15 and the absolute difference between the value of the Blue component of
the pixel with coordinates ( x − 2, y − 2) and the value of the Blue component of the pixel with
coordinates ( x − 1, y − 1) is also greater than 15 than a certain transformation will be applied
to the Blue component of the pixel having the coordinates ( x − 2, y − 2) (assuming x and y are
greater than 2).
arately for the other two components, i.e., Red and Green of each pixel in the image, except for
the pixels in the right and botom edges of the image (those will not be iltered). This remark