Extract handwritten part from the Document

Rahul Kaushik

how do I separate the handwritten part from the pdf in which it has handwritten text as well as printed text. I just want to extract out the handwritten part and save it as another pdf or image.

Suppose there is a pdf/image. Need to extract the Handwritten part from that and save it as another pdf/image.

4 Replies

Know the answer? Post it — somebody with the same question will find it here.

Kip Hackman6y ago
You can use LEADTOOLS ICR SDK technology in your application.
https://www.leadtools.com/sdk/ocr/icr
Here is some sample code:
using (IOcrEngine ocrEngine = OcrEngineManager.CreateEngine(OcrEngineType.OmniPage, false))
using (IOcrDocument ocrDocument = ocrEngine.DocumentManager.CreateDocument())
{
    // Startup the OCR Engine
    ocrEngine.Startup(null, null, null, @"C:\LEADTOOLS 20\Bin\Common\OcrOmniPageRuntime64");
    // Load in image that has handwritten text
    RasterImage image = ocrEngine.RasterCodecsInstance.Load(@"IMAGE FILE PATH", 0, CodecsLoadByteOrder.Rgb, 1, -1);
    // Add this image to the document
    IOcrPage ocrPage = ocrDocument.Pages.AddPage(image, null);
    // Perform default AutoZoning on the page
    ocrDocument.Pages.AutoZone(null);
    // Iterate through each zone in the document to make them ICR zones
    for (int i = 0; i < ocrPage.Zones.Count; i++)
    {
        OcrZone zone = ocrPage.Zones[i];
        zone.ZoneType = OcrZoneType.Icr;
    }
    //Recognize page and output recognition results
    ocrDocument.Pages.Recognize(null);
    ocrDocument.Save(@"OUTPUT RESULTS FILE PATH", DocumentFormat.Text, null);
}
Sundaram Subramanian9y ago
public static void CropDocument(string file, string oldchar, string repChar)
{
int pageNumber = 1;
PdfReader reader = new PdfReader(file);
iTextSharp.text.Rectangle size = new iTextSharp.text.Rectangle(
Globals.fX,
Globals.fY,
Globals.fWidth,
Globals.fHeight);
Document document = new Document(size);
PdfWriter writer = PdfWriter.GetInstance(document,
new FileStream(file.Replace(oldchar, repChar),
FileMode.Create, FileAccess.Write));
document.Open();
PdfContentByte cb = writer.DirectContent;
document.NewPage();
PdfImportedPage page = writer.GetImportedPage(reader,
pageNumber);
cb.AddTemplate(page, 0, 0);
document.Close();
}

Try the above. But only manually able to do. Hope this may helps.
Rahul Kaushik9y ago
Thanks Sundar for the reply. Your some points helped me.
But the challenge or problem i am facing is,
just to crop that part or mark that part which is handwritten

I even don't want to recogonise what it means. I just want to crop that handwritten part
Sundaram Subramanian9y ago
Well, that's actually a pretty complicated task. I have tried something similar before and the results are not always great; however, the program I created, and currently have, does what I need it to do well enough. The reason it worked well for me is due to the fact that I focused on a very specific area of the page, where I knew I would find specifically-formatted text. Mind you, that only worked well for printed text... stamped or hand-written text was a pain to deal with. With enough OCR training, I'm sure that could be fixed, but I didn't have have additional time to devote to that project.
Your results will heavily depend on the OCR method you choose, scan quality, whether it's typed or written by hand, whether the scan is perfectly aligned or slightly skewed, etc., etc., etc.
I'm not going to give you any code, as you won't learn anything if I do, but I'll give you a few tips of how to get started. If you get stuck, post a specific question here and hopefully you'll get some help.
There are many ways and the way I've tried is through conversion of the scanned PDF file into images (one per page). Then, I ran the images through a recognition algorithm in attempt to retrieve all of the text (in my case, it was a specific rectangle on the image).
So, to get images out a PDF, you can use something like Magick.NET. It's even available through NuGet, so that should be an easy part. Since it's a scanned PDF, you should not have any issues getting the images out. There are plenty of tutorials and if you get stuck post a specific question on this site.
The optical character recognition part is the hard part; however, there are some libraries that may assist you or, at the very least, get you started. I used the tessnet2 library (http://www.pixel-technology.com/freeware/tessnet2/).
There are some C# wrappers available and you may even be able to find something from NuGet. Here's one place on GitHub (https://github.com/charlesw/tesseract). Also, have a look here: https://code.google.com/p/tesseractdotnet/ and here: https://github.com/charlesw/tesseract-ocr-dotnet.
Some of those are for tessnet3 and some are for tessnet2. I've had success with tessnet2 32-bit version, but not the others. So, give all of them a go and see what works for you,