public async Task <string> GetText(string tessdataLocation = null)
{
if (!_loaded)
{
await Load();
}
var attempt = GetTextFromSimplePdf();
if (!string.IsNullOrEmpty(attempt) && !string.IsNullOrEmpty(attempt.Trim()))
{
return(attempt);
}
var tiff = await GetImage(ImageFormat.Tiff);
try
{
return(await Tesseract.GetTextFromTiff(tiff, tessdataLocation)); // ocr tiff
}
catch (IOException)
{
tiff = ImageManipulation.ChangeFormat(tiff, ImageFormat.Tiff); // re-saves the image
return(await Tesseract.GetTextFromTiff(tiff, tessdataLocation)); // ocr tiff
}
}
private async Task <byte[]> GetImage(ImageFormat format, bool readAsOneImage = true)
{
if (!_loaded)
{
await Load();
}
return(await Task.Run(() =>
{
_stream.Position = 0;
if (readAsOneImage)
{
var settings = new MagickReadSettings {
Density = new Density(300)
};
using (var images = new MagickImageCollection())
{
images.Read(_stream, settings);
using (var vertical = images.AppendVertically())
{
var path = $"pdf.{GetFilenameExtension(format)}";
vertical.Write(path); // todo handle with stream instead of file
return File.ReadAllBytes(path);
}
}
}
else
{
var settings = new MagickReadSettings {
Density = new Density(300, 300)
};
var pages = new List <byte[]>();
using (var images = new MagickImageCollection())
{
images.Read(_stream, settings);
var page = 1;
foreach (var image in images.OfType <MagickImage>())
{
var path = $"pdf.{page}.{GetFilenameExtension(format)}";
image.Write(path);
pages.Add(File.ReadAllBytes(path));
page++;
}
}
return ImageManipulation.CombineImages(pages, format);
}
}));
}