Extracting text from animated GIFs

[Daniel Bishop]

Hello everyone!

I'm just getting started with Tesseract and am wowed at how well it does on tasks like scanned black and white text! I'm... less than thrilled at how it does at my current endeavor, which is to extract the text from animated GIFs, such as from reaction GIFs and memes and so on.

After reading the FAQ and the ImproveQuality articles as well as some further prodding around, it seems the DPI of the images isn't too small, but rather that most of the issue comes from the variety of background colors around the text and/or the font(s) commonly used for memes.

Does anyone have any experience with this, or have any helpful advice for this specific task? Attached is a sample of the kind of thing I want to process.

Thank you for your time.

(Incidentally, even though the new version of leptonica and tesseract both say they support gifs, I get the following error when I send a gif in:

Tesseract Open Source OCR Engine v3.04.00 with Leptonica

Error in pixReadMemGif: function not present

Error in pixReadMem: gif: no pix returned

Error during processing.

)