I understand that you can edit box files with any editor (even text editor) and check it. Been there, done that, awesome feature.
I'm curious if it is possible to feed tesseract predefined boxes for it to just use OCR inside ? I'll make sure that all the boxes have only one character inside, promise. Or should I just remove everything else, move character regions a little apart etc. to fool the system ? That would be really heavy load, would rather just provide boxes to the system. Also, I'm curious for the tesseract methods on different parts. If they are not confidential, of course. 1-) Image Enhancement Do you use any image enhancement ? Contrast enhancement ? Histogram equalization ? Anything ? 2-) Text Detection (as in finding lines / words / candidates) Does it use edge detectors? Which one(s)? Does it use basic dynamic thresholding with a mask? How does it determine mask sizes? 3-) Character Segmentation / Validation Does it use connected component analysis? Projection profile? Something else? 4-) Character Recognition. k-means to get cluster centroids ? knn ? svm ? mlp? 5-) Word Validation Don't suppose you are using a dictionary, yet. So don't think this part is in the tesseract. Therefore, if you could return best two matches (if they have at least %60 of the distance to respective centroids) for me to use validation on dictionary. If you could explain any tiny part, I'd really appreciate it. Cheers, Cihan -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/tesseract-ocr?hl=en
