meh / ruby-tesseract-ocr

A Ruby wrapper library to the tesseract-ocr API.
629 stars 74 forks source link
ruby rubynlp tesseract-ocr wrapper

ruby-tesseract - Ruby bindings and wrapper

This wrapper binds the TessBaseAPI object through ffi-inline (which means it will work on JRuby too) and then proceeds to wrap said API in a more ruby-esque Engine class.

Making it work

To make this library work you need tesseract-ocr and leptonica libraries and headers and a C++ compiler.

The gem is called tesseract-ocr.

If you're on a distribution that separates the libraries from headers, remember to install the -dev package.

On Debian you will need to install libleptonica-dev and libtesseract-dev.

Examples

Following are some examples that show the functionalities provided by tesseract-ocr.

Basic functionality of tesseract

require 'tesseract'

e = Tesseract::Engine.new {|e|
  e.language  = :eng
  e.blacklist = '|'
}

e.text_for('test/first.png').strip # => 'ABC'

You can pass to #text_for either a path, an IO object, a string containing the image or an object that responds to #to_blob (for example Magick::Image), keep in mind that the format has to be supported by leptonica.

Accessing advanced features

With advanced features you get access to blocks, paragraphs, lines, words and symbols.

Replace level in method names with either block, paragraph, line, word or symbol.

The following kind of accessors need a block to be passed and they pass to the block each Element object. The Element object has various getters to access certain features, I'll talk about them later.

The methods are:

The following accessors instead return an Array of Elements with cached getters, the getters are cached beacause the values accessible in the Element are linked to the state of the internal API, and that state changes if you access something else.

The methods are:

Again, to *_for methods you can pass what you can pass to a #text_for.

Each Element object has the following getters:

Block elements also have type accessors that specify the type of the block.

Word elements also have font_attributes, from_dictionary? and numeric? getters.

Symbol elements also have superscript?, subscript? and dropcap? getters.

hOCR

require 'tesseract'

e = Tesseract::Engine.new {|e|
  e.language  = :eng
  e.blacklist = '|'
}

puts e.hocr_for('test/first.png')

You can pass to #hocr_for either a path, an IO object, a string containing the image or an object that responds to #to_blob (for example Magick::Image), keep in mind that the format has to be supported by leptonica.

Please note you have to pass #hocr_for the page you want to get the output of as well.

Using the binary

You can also use the shipped executable in the following way:

> tesseract.rb -h
Usage: tesseract [options]
        --path PATH                  datapath to set
    -l, --language LANGUAGE          language to use
    -m, --mode MODE                  mode to use
    -p, --psm MODE                   page segmentation mode to use
    -u, --unlv                       output in UNLV format
    -c, --confidence                 output the mean confidence of the recognition
    -C, --config PATH...             config files to load
    -b, --blacklist LIST             blacklist the following chars
    -w, --whitelist LIST             whitelist the following chars
> tesseract.rb test/first.png 
ABC
> tesseract.rb -c test/first.png 
86

License

The license is BSD one clause.