Hi Andrew,
your thought is exactly what I would do: replace RGB (201, 201, 201) with
RGB (255, 255, 255) and then process the image with text on white ground. :
)
Ale
Il giorno domenica 11 dicembre 2022 alle 08:48:40 UTC+1 Andrew ha scritto:
> I have processed imaged that seems pretty simple:
run this to understand what it problem:
> tesseract 8fXlqZY.png 8fXlqZY --psm 7 get.images
Then check the binarized version of your input that is used for OCR:
8fXlqZY.processed.tif
There are 2 simple ways to solve the problem:
1. using only text ares for OCR (e.g. cropping image to text wit
2 matches
Mail list logo
