What is segmentation and how is it used?
Segmentation is the process of taking an image and splitting it up into smaller bitmaps. Each of these smaller bitmaps is supposed to contain the pixels put there by a single character or glyph. The first step is to do a horizontal pixel histogram (sum the pixels in each row) and look for minima (or runs of scan lines where the pixel count is below a given threshold). The next step is to do the same for columns of pixels and look for stretches where the vertical histogram values are below a given threshold. The result of this is a set of “regions of interest”. The coordinates are determined by the starting and ending lines and columns at which the horizontal and vertical pixel histograms are above threshold. Segmentation occurs early in the processing of ICR and OCR fields. Segmentation as discussed here is not performed for OMR fields. “Do multiple lines” – means do a horizontal pixel histogram to find multiple lines in the field. If this is set, then “minimum line separation pixels”
