sk-spell

podpora slovenčiny v Open Source programoch

Tesseract Project Dependencies   

back to tesseract-ocr-en

A long term aim of tesseract project is to keep dependency on external libraries as low as possible. Anyway more features usually means that usage of external libraries will be increasing instead of implementation of own solution (especially for training tools). If you plan to use tesseract in your own application it is question what you really needs.

On some platforms (e.g. Linux) it is not problem – tesseract use system wide libraries. But on other platforms (e.g. Windows) you can have huge problems if you want to build everything by yourself. Here is short overview about dependency on (main) libraries.

I want to test building of tesseract with clang on windows from scratch so I have to solve this. This article is description of my findings. Maybe it is not full because some programs&libraries I already have installed without recognizing them as tesseract dependency.

Tesseract 4.0 code can be divided to following parts:

1. tesseract executable (example of usage tesseract library)
2. tesseract library
3. tesseract training tools

Tesseract Executable Dependencies

• Tesseract library (Required) – for OCR
• Leptonica (Required) – for reading images
• TIFF (Optional) – just for handling tiff error messages

Tesseract Library Dependencies

• OpenCl (Optional) – still experimental and not maintained with in tesseract
• Leptonica (Required) – for image operation
  • zlib (Optional)
  • PNG | libpng (Optional)
    • requires zlib
  • GIF | GIFLIB (Optional)
  • JPEG (Optional)
  • TIFF | LibTIFF (Optional)
    • zlib
    • JPEG
    • JBIG
    • LibLZMA | XZ
    • GLUT
    • OpenGL
    • ZSTD
    • WEBP
  • openjpeg/libopenjp2 (Optional):
    • PNG
    • TIFF
    • LCMS2
    • LCMS
  • WEBP (Optional)
    • PNG
    • JPEG
    • TIFF
    • GIF
    • GLUT
    • SDL

Tesseract Training Tools Dependencies

This is not fully tested, I just grab it from cppan log:

• Tesseract library (Required)
• pango (Required)
• cairo (Required)
• icu (Required)
• pangocairo
• cairographics
• fontconfig
• expat
• freetype
• bzip2
• pixman
• glib
• pangoft2
• gobject
• libffi
• gettext
• iconv
• pcre
• fribidi
• harfbuzz

Conclusion

My understanding is that when tesseract is used with other project like QT or OpenCV (that has solution for opening images), there is no need to compile leptonica with support of various image types, just to convert (e.g. QImage or mat) image to PIX structure. This can simplify building tesseract library significantly.

back to tesseract-ocr-en

© projekt sk-spell