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.
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
.
Following are some examples that show the functionalities provided by tesseract-ocr.
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.
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:
each_level
each_level_for
each_level_at
The following accessors instead return an Array
of Element
s 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:
levels
levels_for
levels_at
Again, to *_for
methods you can pass what you can pass to a #text_for
.
Each Element
object has the following getters:
bounding_box
, this will return the box where the element is confined intobinary_image
, this will return the bichromatic image of the elementimage
, this will return the image of the elementbaseline
, this will return the line where the text is with a pair of
coordinatesorientation
, this will return the orientation of the elementtext
, this will return the text of the elementconfidence
, this will return the confidence of correctness for the elementBlock
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.
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.
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
The license is BSD one clause.