org.safs.tools.ocr.tesseract

Class TesseractOCREngine

java.lang.Object

org.safs.tools.ocr.OCREngine

org.safs.tools.ocr.tesseract.TesseractOCREngine

public class TesseractOCREngine
extends OCREngine

Extends OCREngine to support Tesseract OCR engine. http://code.google.com/p/tesseract-ocr/ http://groups.google.com/group/tesseract-ocr Tesseract 2.04 provides two ways for using 1) Running tesseract.exe in command line directly. Usage: tesseract.exe [-l lang] [configfile] 2) Tesseract releases tessdll.dll for developers to call, dlltest.exe is released for testing tessdll.dll. Files tesseract.exe, tessdll.dll and dlltest.exe can be downloaded, or generated from the released code. SAFS used to take the second way -- SAFS calls tessdllWrapper.dll(SAFS defined), which talks with tessdll.dll. Some experiments showed tesseract.exe seemed much better than tessdll.dll in detecting text on images. Not sure if they use the same parts of code to do the work. Just take the first way -- running tesseract.exe directly. Two files will be output in current user directory if call imageToText() 1. ~temp.tif scaled image for Tesseract to recognize. 2. ~temp.txt text file storing detected text in the image. Three files required in searching path: tesseract.exe --- Command: tesseract imagefile outfile -l eng SafsTessdll.exe --- Command: SafsTessdll imagefile outfile eng tessdll.dll --- needed by SafsTessdll.exe SafsTessdll.exe was built by SAFS and newly added for findTextRectFromImage(). It outputs a UTF-8 file that contains detected character, their Unicode and their coordinates.

We have added direct tesseract.exe support for findTextRectFromImage(). The output format of this file is different and uses a different coordinate system then the SAFS DLL, but it provides greater text recognition accuracy for better matching and locating text.

If a version other than Tesseract 2.04 is installed, the environment variable TESSDATA_VERSION should be set. Ex: TESSDATA_VERSION=3.4.4
We don't know of any other means to deduce the version of tesseract installed.

Author:

Junwu Ma
DEC 14, 2009 Original Release
JAN 27, 2010 (JunwuMa) Modified imageToText() to call tesseract.exe directly for recognition accuracy.
FEB 25, 2010 (JunwuMa) Add method getSelfDefinedLangId().
MAR 19, 2010 (JunwuMa) Added method findTextRectFromImage() to support mode "ImageText=" in ImageUtils.java.
MAR 25, 2010 (JunwuMa) Modified imageToText() to open the temporary file in proper format, UTF-8.
APR 20, 2010 (Lei Wang) Add a map languages to contain pairs (javaLangCode, OCRLangCode). Modify method getSelfDefinedLangId(): get OCRLangCode from map languages.
MAY 27, 2010 (Carl Nagle) changed temp output directory to System property "java.io.tmpdir"
OCT 22, 2010 (Carl Nagle) added support for tesseract.exe text coordinate extraction added support to detect System Environment Variable TESSDATA_VERSION to detect versions of Tesseract > 2.04.

See Also:

tessFileParser, ReverseRectangle

Field Summary

Fields

Modifier and Type	Field and Description
static java.lang.String	LANG_CHN "chi"
static java.lang.String	LANG_ENG "eng"
static java.lang.String	LANG_FRA "fra"
static java.lang.String	LANG_JPN "jpn"
static java.lang.String	LANG_KOR "kor"
static java.util.HashMap<java.lang.String,java.lang.String>	languages
static java.lang.String	TESSERACT_VERSION Coded default to "2.04" Static initializers look for System Environment variable "TESSDATA_VERSION" to change this.
static int	TEXT_FIND_MODE Set to desired mode for locating coordinates of text. TEXT_FIND_TESSDLL_MODE was the original mechanism used.
static int	TEXT_FIND_TESSDLL_MODE Mode value to use SAFSTESSDLL wrapper to locate coordinates of image text.
static int	TEXT_FIND_TESSEXE_MODE Mode value to use tesseract.exe directory to locate coordinates of image text.
static java.lang.String	TMP_DIR_PROPERTY "java.io.tmpdir"
static java.lang.String	TMP_TEXT_COOR_OUTPUT "~tempcoor.txt"
static java.lang.String	TMP_TEXT_COOR_ROOT "~tempcoor"
static java.lang.String	TMP_TEXT_OUTPUT "~temp"
static java.lang.String	TMP_TIF_SCALEDED "~temp.tif"

Fields inherited from class org.safs.tools.ocr.OCREngine

defaultZoomScale, OCR_DEFAULT_ENGINE_CLAZZ, OCR_DEFAULT_ENGINE_KEY, OCR_G_ENGINE_CLAZZ, OCR_G_ENGINE_KEY, OCR_T_ENGINE_CLAZZ, OCR_T_ENGINE_KEY, STAF_OCR_ENGINE_VAR_NAME, STAF_OCR_LANGUAGE_ID_VAR_NAME

Constructor Summary

Constructors

Constructor and Description
TesseractOCREngine()

Method Summary

Modifier and Type	Method and Description
java.awt.Rectangle	findTextRectFromImage(java.lang.String searchtext, int index, java.awt.image.BufferedImage image, java.lang.String stdlangId, java.awt.Rectangle subarea, float zoom) Find text from BufferedImage, and return its area in the BufferedImage.
protected java.lang.String	getSelfDefinedLangId(java.lang.String StdlangId) Override its super.
java.lang.String	imageToText(java.awt.image.BufferedImage image, java.lang.String langId, java.awt.Rectangle subarea, float zoom) Convert buffered image to text using OCR technology.
static void	main(java.lang.String[] args) Can be used to unit test.

Methods inherited from class org.safs.tools.ocr.OCREngine

getdefaultZoomScale, getOCREngine, getOCREngineKey, getOCRLanguageCode, imageToText, runCommandLine, setdefaultZoomScale, setOCREngineKey, setOCRLanguageCode, storedImageToText, storedImageToText, zoomImageWithType

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

LANG_ENG

public static final java.lang.String LANG_ENG

"eng"

See Also:

Constant Field Values

LANG_CHN

public static final java.lang.String LANG_CHN

"chi"

See Also:

Constant Field Values

LANG_JPN

public static final java.lang.String LANG_JPN

"jpn"

See Also:

Constant Field Values

LANG_KOR

public static final java.lang.String LANG_KOR

"kor"

See Also:

Constant Field Values

LANG_FRA

public static final java.lang.String LANG_FRA

"fra"

See Also:

Constant Field Values

TMP_DIR_PROPERTY

public static java.lang.String TMP_DIR_PROPERTY

"java.io.tmpdir"

TMP_TIF_SCALEDED

public static java.lang.String TMP_TIF_SCALEDED

"~temp.tif"

TMP_TEXT_OUTPUT

public static java.lang.String TMP_TEXT_OUTPUT

"~temp"

TMP_TEXT_COOR_ROOT

public static java.lang.String TMP_TEXT_COOR_ROOT

"~tempcoor"

TMP_TEXT_COOR_OUTPUT

public static java.lang.String TMP_TEXT_COOR_OUTPUT

"~tempcoor.txt"

TESSERACT_VERSION

public static java.lang.String TESSERACT_VERSION

Coded default to "2.04" Static initializers look for System Environment variable "TESSDATA_VERSION" to change this. We have not found another way to deduce the installed version of Tesseract.

languages

public static java.util.HashMap<java.lang.String,java.lang.String> languages

TEXT_FIND_TESSDLL_MODE

public static final int TEXT_FIND_TESSDLL_MODE

Mode value to use SAFSTESSDLL wrapper to locate coordinates of image text.

See Also:

Constant Field Values

TEXT_FIND_TESSEXE_MODE

public static final int TEXT_FIND_TESSEXE_MODE

Mode value to use tesseract.exe directory to locate coordinates of image text.

See Also:

Constant Field Values

TEXT_FIND_MODE

public static int TEXT_FIND_MODE

Set to desired mode for locating coordinates of text.
TEXT_FIND_TESSDLL_MODE was the original mechanism used. However, this mechanism was found to lack the accuracy of text recognition that tesseract.exe has.

Possible values: TEXT_FIND_TESSDLL_MODE, and the default TEXT_FIND_TESSEXE_MODE.

Constructor Detail

TesseractOCREngine

public TesseractOCREngine()

Method Detail

imageToText

public java.lang.String imageToText(java.awt.image.BufferedImage image,
                                    java.lang.String langId,
                                    java.awt.Rectangle subarea,
                                    float zoom)
                             throws SAFSException

Description copied from class: OCREngine

Convert buffered image to text using OCR technology. It needs to be implemented in its derived class.

Overrides:

imageToText in class OCREngine

Returns:

String, converted from the input image. NULL if fails to convert.

Throws:

SAFSException - if meets any Exception

getSelfDefinedLangId

protected java.lang.String getSelfDefinedLangId(java.lang.String StdlangId)

Override its super. Translate the standard language code to OCR specific language code. It should be overridden in derived classes. Refer to Locale.ENGLISH.getLanguage() for input langid. For example: 'en' -- English

Overrides:

getSelfDefinedLangId in class OCREngine

Parameters:

langId, - standard language code

Returns:

findTextRectFromImage

public java.awt.Rectangle findTextRectFromImage(java.lang.String searchtext,
                                                int index,
                                                java.awt.image.BufferedImage image,
                                                java.lang.String stdlangId,
                                                java.awt.Rectangle subarea,
                                                float zoom)
                                         throws SAFSException

Find text from BufferedImage, and return its area in the BufferedImage. Two modes of operation are possible based on the TEXT_FIND_MODE setting.

 TESSDLL Mode:
 Two files SafsTessdll.exe and tessdll.dll will be used.
 Two files will be output in current user's Temp directory.
 1. ~temp.tif  scaled image
 2. ~tempcoor.txt text file storing detected text and their coordinates.
    Rectangle coordinates 0,0 relative to TOP-LEFT corner of search area.
 

 TESSEXE Mode:
 tesseract.exe will be used.
 Two files will be output in current user's Temp directory.
 1. ~temp.tif  scaled image
 2. ~tempcoor.txt text file storing detected text and their coordinates.
    ReverseRectangle coordinates 0,0 relative to BOTTOM-LEFT corner of search area.
 

Overrides:

findTextRectFromImage in class OCREngine

Parameters:

searchtext, - text for which to search

index, - starts from 1, specifies to find the Nth instance of searchText.

image, - source BufferedImage for detecting

stdlangId, - standard language id with which TOCR intends to detect. Refer to Locale.ENGLISH.getLanguage().

subarea -

zoom -

Returns:

Rectangle or ReverseRectangle or null.

Throws:

SAFSException

See Also:

#TEXT_FIND_MODE}, ReverseRectangle

main

public static void main(java.lang.String[] args)

Can be used to unit test.

java org.safs.tools.ocr.tesseract.TesseractOCREngine imageFile [-z zoom] [-l lang] [-t text]

imageFile - path to screen captured image to process.
zoom - zoom level to use for OCR. Defaults to 1.9.
lang - locale to use for OCR. Ex: "en"
text - text to locate in image.

Parameters:

args - -- up to 7 array items: imageFile [-z zoom] [-l lang] [-t text] imageFile - (required) path to screen captured image to process.
zoom - zoom level to use for OCR. Defaults to 1.9.
lang - locale to use for OCR. Ex: "en". Defaults to "en".
text - text to locate in image.