Image Processing Reference
In-Depth Information
As a remark, the same pixel could be the result of applying two types of transformations to
its components (e.g., Red, Green, or Blue). For instance, we could focus on an arbitrary pixel
that has the coordinates ( x , y ) relative to the upper-left corner of the image. By assuming that
Equation (3) is applied to the Red component of that pixel and Equation (4) is applied to the
Blue component of the same pixel, we are facing a possible situation that could arise in the
algorithm. Despite this fact exemplified before, not more than one transformation will be ap-
plied to a single component of a pixel (e.g., Red component could not possibly be the result of
applying Equations (3) and (4) , it will have to be either (3) or (4) , but not both).
Furthermore, two different pixels could be the result of applying different transformations
to the same component of the two pixels, e.g., the filtered Red component of the first pixel
that has the coordinates ( x 1 , y 1 ) could be the result of applying Equation (3) and the filtered
Red component of the second pixel that has the coordinates ( x 2 , y 2 ) could be the result of ap-
plying Equation (4) . The graphic representation of the two transformations is shown in Figure
4(a) , together with the identity function that helps in spoting the way pixel components are
increased or decreased. Figure 4(b) highlights the difference between Contrast and Smart Con-
trast algorithms. If the value of the pixel component is less than threshold 127, the blue plot
describes the transformation that is applied to that certain component, else the transformation
shown in the red plot is the one applied to the component.
FIGURE 4 (a) Smart Contrast transformation (functions f and g) and the identity function h.
(b) Smart Contrast (functions f and g) and contrast transformation (function h).
Since the one and the same transformation is not applied to all pixels in the image, the “color
matrix” optimization techniques cannot be used when it comes to Smart Contrast image filter
Another optimization technique is applied to this filter, as well as to the further image filter,
namely, Highlight, and will be detailed at the end of the description of both image ilters.
Best OCR rate of success for the filtered images using Smart Contrast filter is produced
when contrast scale is set to 1.5.
2.3 Visual Result of Applying Smart Contrast on Images
Smart Contrast filter produces the results shown in Figure 5 . The results produced by Contrast
filter are also shown in Figure 5 in order to spot the improvements made by Smart Contrast.
The effect of applying Smart Contrast filter would be that, in most of the cases, contrast is in-
creased in areas where edges of the objects (e.g., characters) appear in the image. Exceptions
occur when the color of the characters is close to the color of the background, but both are
close to either the lowest color intensity or the highest color intensity. This drawback is solved
in Highlight filter.
FIGURE 5 The visual effect of Smart Contrast image filter.
 
 
Search WWH ::




Custom Search