On Monday, July 13, 2015 at 1:23:10 AM UTC-4, Elan wrote: > > > I am fairly new to tesseract, I have done some playing around with > training new fonts, and loading config files etc. I have an issue with the > images I am trying to OCR. > In many cases, there is a dotted horizontal line about 5-10 pixels above > the text. Tesseract mistakenly assumes this is apart of the text and puts > the box around the character and around the line above the text. > An example below > > _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Example of > lines above text > > 1. text to read > 2. text to read > 3. text to read > 4. text to read > _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ > > It reads lines 2 and 3 almost perfectly, however, 1 and 4 is inconsistent > and can vary. Most of the time its gibberish. It makes it hard to train > tesseract properly as the box files have been produced > > I was wondering if there is a parameter or configuration were I could set > the maximum font size or maximum box size to avoid it from including the > lines above the text? > I would do some morphological operations on the lines to get rid of them > but the lines are about the same thickness as the font and I would worry it > would degrade the text. > I know tesseract requires minimum size 10 font to get acceptable results, > so I was wondering if there is a way to set the max font size. > > The font size should be fairly even across the images (obviously camera > distortion may result in an offset of a pixel or two but roughly the same) > > I am aware I could segment the image and pull out the regions in between > the lines. I guess I am just seeing if there is a quick configuration or > parameter I could parse to satisfy this requirement? > > Can anyone help me? > Is pre-processing the only way to solve this? >
I'm confused by your description. Is this for training? If so, there are probably not very many images and you can just edit them by hand. If this is not training, but regular recognition, you might want to investigate the page segmentation mode parameter. Tom -- 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/dd14fd21-2743-46af-bb46-df86acc3a205%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
