benckx / optimize-pdf-ereaders

Optimize scanned PDFs for small ebook readers using OCR

About

PDF books and articles found online are usually poorly rendered on small e-readers (e.g. Kindle Oasis), as a whole PDF page is displayed on the small screen.

This lib uses OCR to correct the skewed angle of the page, crop around the text and re-paginate; as to optimize for the best reading experience on small e-readers.

The code was initially written in 2018 in Java, alongside an online converter website that I decided to take down as it would cost quite a bit (OCR and image processing being quite resource-intensive). I also couldn't maintain it as I was working full time.

Therefore, the project probably needs a bit of a cleanup.

The unit tests using full PDF books can not be shared publicly, so I will re-add them later, using only individual pages rather than complete books.

Examples

Example 1

Input

download PDF

Output

download PDF

Example 2

Input

download PDF

Output

download PDF

Example 3

Input

download PDF

Output

download PDF

Requirements

sudo apt-get install tesseract-ocr

The data in tessdata/ is found on https://github.com/tesseract-ocr/tessdata_best

Usage

    RequestConfig requestConfig = RequestConfig
        .builder()
        .pdfFile(file)
        .minPage(minPage)
        .maxPage(maxPage)
        .correctAngle(true)
        .build();

    Processor processor = new Processor(requestConfig);
    processor.process();
    processor.joinThread();
    File outputFile = processor.writeToPDFFile(fileName + "_optimized.pdf");

TODO

• Move to Gradle
• Re-add unit tests that can be shared publicly, adapt the other ones
• Add language as a parameter
• Create a user-friendly runnable
• Move to Kotlin
• Finish picture detection