EN
152
Software Module
11.15.2.1. Examples
Ideal edge spacing
Reading time: 25 ms
Edge spacing too large: The segmentation function
detects additional object because an incorrect
binarization threshold is calculated due to the large
surface area of the image.
Reading time: 120 ms
11.15.3. Background
A homogenous background is always ideal for character segmentation. Structures included in the background
which have an intensity similar to that of the characters make reading impossible.
Under certain circumstances, colored structures can be eliminated by using the right illumination color. If fine
structures are present in the background, it may be helpful to make use of a Gaussian filter or set the optics
slightly out of focus depending upon character size.
Character string with structure in background
segmentation is not possible.
Character string with structure in background
segmentation functions correctly.
11.15.4. Contrast
The “binarization” stage must find a suitable binarization threshold for separating the characters from the
background on the basis of image contrast. The OCR module is equipped with various binarization functions
to this end.
If contrast (difference in intensity between characters and background) is constant over the entire ROI, 20
intensity values are enough for display.
If contrast varies within the ROI (e.g. due to inhomogeneous illumination), there should be a plain difference
between the characters and the background. The sensor functions ideally if the image is set up with black
characters (intensity = 0) and a gray background. In this case, brightness differences within the characters are
outside of the image chip’s dynamic range, and the characters are entirely black.
On the other hand, an attempt can be made to fully over-illuminate the background (white, i.e. intensity = ma-
ximum). In this way, structures in the background can no longer be detected and only the characters are gray.
To Next Page
Loading...