I am using Tesseract 3.04 on Windows to analyze scanned paper forms which often contain non-contiguous text labels of various size, position, and font style. I am attempting to deduce simple typeface characteristics such as serif vs sans-serif, fixed vs variable pitch, italics, bold, etc, in an effort to loosely classify identified text labels.
I started out by using LTRResultIterator::WordFontAttributes for recognized words, but then learned that the returned font properties are from the *best* matching font, not from an accumulation of actual character attributes for the recognized word. As an example of this, I have observed cases where sequences of ARIAL (sans-serif, variable-pitch) characters are measured and determined to be fixed-pitch (for example: "BOOK"), and the best matching font is a COURIER variant (fixed-pitch, serif). In this case, none of the characters have serifs, but the determined pitch (fixed) seems to carry significant weight when matching fonts. I intend to study the font classification logic a bit to be sure I understand it. I also suspect that the Adaptive Classifier may propagate this effect for 'downstream' results. (True/False? Opinions?) I thought about exploring the following: 1. disabling fixed-pitch character classification and handling 2. disabling the adaptive classifier or limiting it's influence Does anyone have any suggestions or opinions? Thanks, Vince Roscigno -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/tesseract-ocr. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/3cc3f26e-ad8e-4a81-995f-2b1b0c14d0db%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
