/// <summary>
/// Check whether tesseract executable is installed on the machine and
/// provided path to tesseract executable is correct.
/// </summary>
/// <param name="execPath">path to tesseract executable</param>
private void CheckTesseractInstalled(String execPath)
{
try {
TesseractHelper.RunCommand(execPath, JavaCollectionsUtil.SingletonList <String>("--version"));
}
catch (Tesseract4OcrException e) {
throw new Tesseract4OcrException(Tesseract4OcrException.TESSERACT_NOT_FOUND, e);
}
}
/// <summary>
/// Performs tesseract OCR using command line tool for the selected page
/// of input image (by default 1st).
/// </summary>
/// <remarks>
/// Performs tesseract OCR using command line tool for the selected page
/// of input image (by default 1st).
/// Please note that list of output files is accepted instead of a single file because
/// page number parameter is not respected in case of TIFF images not requiring preprocessing.
/// In other words, if the passed image is the TIFF image and according to the
/// <see cref="Tesseract4OcrEngineProperties"/>
/// no preprocessing is needed, each page of the TIFF image is OCRed and the number of output files in the list
/// is expected to be same as number of pages in the image, otherwise, only one file is expected
/// </remarks>
/// <param name="inputImage">
/// input image
/// <see cref="System.IO.FileInfo"/>
/// </param>
/// <param name="outputFiles">
///
/// <see cref="System.Collections.IList{E}"/>
/// of output files
/// (one per each page)
/// </param>
/// <param name="outputFormat">
/// selected
/// <see cref="OutputFormat"/>
/// for tesseract
/// </param>
/// <param name="pageNumber">number of page to be processed</param>
internal override void DoTesseractOcr(FileInfo inputImage, IList <FileInfo> outputFiles, OutputFormat outputFormat
, int pageNumber)
{
ScheduledCheck();
IList <String> @params = new List <String>();
String execPath = null;
String imagePath = null;
try {
imagePath = inputImage.FullName;
// path to tesseract executable
if (GetPathToExecutable() == null || String.IsNullOrEmpty(GetPathToExecutable()))
{
throw new Tesseract4OcrException(Tesseract4OcrException.CANNOT_FIND_PATH_TO_TESSERACT_EXECUTABLE);
}
else
{
if (IsWindows())
{
execPath = AddQuotes(GetPathToExecutable());
}
else
{
execPath = GetPathToExecutable();
}
@params.Add(execPath);
}
CheckTesseractInstalled(execPath);
// path to tess data
AddTessData(@params);
// validate languages before preprocessing started
ValidateLanguages(GetTesseract4OcrEngineProperties().GetLanguages());
// preprocess input file if needed and add it
imagePath = PreprocessImage(inputImage, pageNumber);
AddInputFile(@params, imagePath);
// move to image directory as tesseract cannot parse non ascii
// characters in input path
IList <String> moveToDirectoryParams = MoveToImageDirectory(imagePath);
// output file
AddOutputFile(@params, outputFiles[0], outputFormat, imagePath);
// page segmentation mode
AddPageSegMode(@params);
// add user words if needed
AddUserWords(@params, imagePath);
// required languages
AddLanguages(@params);
if (outputFormat.Equals(OutputFormat.HOCR))
{
// path to hocr script
SetHocrOutput(@params);
}
// set default user defined dpi
AddDefaultDpi(@params);
OnEvent();
TesseractHelper.RunCommand(IsWindows() ? "cmd" : "bash", CreateCommandList(moveToDirectoryParams, @params)
);
}
catch (Tesseract4OcrException e) {
LogManager.GetLogger(GetType()).Error(e.Message);
throw new Tesseract4OcrException(e.Message, e);
}
finally {
try {
if (imagePath != null && !inputImage.FullName.Equals(imagePath))
{
TesseractHelper.DeleteFile(imagePath);
}
}
catch (SecurityException e) {
LogManager.GetLogger(GetType()).Error(MessageFormatUtil.Format(Tesseract4LogMessageConstant.CANNOT_DELETE_FILE
, imagePath, e.Message));
}
try {
if (GetTesseract4OcrEngineProperties().GetPathToUserWordsFile() != null && GetTesseract4OcrEngineProperties
().IsUserWordsFileTemporary())
{
TesseractHelper.DeleteFile(GetTesseract4OcrEngineProperties().GetPathToUserWordsFile());
}
}
catch (SecurityException e) {
LogManager.GetLogger(GetType()).Error(MessageFormatUtil.Format(Tesseract4LogMessageConstant.CANNOT_DELETE_FILE
, GetTesseract4OcrEngineProperties().GetPathToUserWordsFile(), e.Message));
}
}
}