I have curiosity about the limitation of tesseract in reading text in image. I have tested tesseract for image like pdf screenshot and I satisfied with the result.
But, when I tried tesseract for text in natural image or image that the text is manually added by user in random background (ex: googling interesting image and adding text in photoshop), tesseract has such difficulties to read it, of course because the variety of background. Therefore, I did some preprocessing like finding text edges, grayscaling images etc to make the image more uniform. Then, I got interesting result. I did 2 preprocessing, grayscaling and doing histogram equalization, and save the image with 2 different names then run tesseract to read it. For the 1st image (grayscaling), tesseract can read 3 of 4 words correctly, and for the 2nd image, tesseract can read nothing. And then, I stack the images side to side and save it as one image, then run tesseract to read it. The result is, for the first image, tesseract can read correctly and for the 2nd image, tesseract can read 3/4 words. So, what kind of image which tesseract can read? note: Im using my own tesseract model in tesseract 4.0 with word_dawg= 0.5 -- 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 https://groups.google.com/group/tesseract-ocr. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/384b606d-0e29-41df-8740-f9a82f040dd4%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
