A lot of people seem to assume that if they get poor quality results from Tesseract the appropriate response is to retrain tesseract. For most cases that will just be a waste of time, and preprocessing would be a much more sensible route.
While this is not news for those of us who have been around for very long here, I think it could be better communicated to new users. There is a good brief entry in the FAQ about it[0], but the FAQ is long, and I think a more focused separate page which is referenced from the homepage would be a better idea. I've drafted such a page, and I'd be keen to get feedback on it. Is it clear? Is it a good idea? I haven't filled out all of the "Image processing" sections yet, but (presuming people don't hate the idea in general) I will do soon, including image examples. The page is attached, as is an image that will be displayed in the "binarisation" section (shamelessly stolen from Zdenko's page at [1]). Nick 0. http://code.google.com/p/tesseract-ocr/wiki/FAQ#Output_without_result_or_bad_output 1. http://www.sk-spell.sk.cx/through-tesseract-ocr-eye -- -- 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 --- 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]. For more options, visit https://groups.google.com/groups/opt_out.
= Poor quality output = There are a variety of reasons you might not get good quality output from Tesseract. It's important to note that unless you're using a very unusual font or a new language retraining Tesseract is unlikely to help. <wiki:toc /> == DPI == Tesseract works best with text using a DPI of at least 300 dpi, so it may be beneficial to resize images. For more information see [FAQ#Is_there_a_Minimum_Text_Size?_%28It_won%27t_read_screen_text!%29 the FAQ]. == Image processing == Tesseract does various image processing operations internally (using the Leptonica library) before doing the actual OCR. It generally does a very good job of this, but there will inevitably be cases where it isn't good enough, which can result in a significant reduction in accuracy. You can see how Tesseract has processed the image by using the configuration variable <code>tessedit_write_images = true</code> when running Tesseract. If the resulting <code>tessinput.tif</code> file looks problematic, try some of these image processing operations before passing the image to Tesseract, either manually using a graphics editor like [http://www.gimp.org Gimp], or in code using an image processing library like [http://leptonica.com Leptonica]. === Binarisation === <img style="float:right; width:50%;" src="binarisation.png" /> This is converting an image to black and white. Tesseract does this internally, but it can make mistakes, particularly if the page background is of uneven darkness. <div style="clear:right"></div> === Noise === === Warp === === Blur / Sharpen === == Segmentation method == By default Tesseract segments an image expecting a page of text. If you're just seeking to OCR a small region try a different segmentation mode, using the <code>-psm</code> argument. == Still having problems? == If you've tried the above and are still getting low accuracy results, [https://groups.google.com/forum/?fromgroups#!forum/tesseract-ocr ask on the forum] for help, ideally posting an example image.
<<attachment: binarisation.png>>
