OCR not reading images created from MemoryStream or Bitmap

I am using the IronOCR to read an image to text, I ave also tried Tesseract and Tessnet to no avail. However, it is not reading the file created from my MemoryStream() or just Bitmap.Save in general.

The code below is funcitonal: 

class Screenshot
{
public static bool saveToClipboard = true;
public static void CaptureImage(bool showCursor, Size curSize, Point curPos, Point SourcePoint, Point DestinationPoint, Rectangle SelectionRectangle, string FilePath, string extension)
{
using (Bitmap bitmap = new Bitmap(SelectionRectangle.Width, SelectionRectangle.Height))
{
using (Graphics g = Graphics.FromImage(bitmap))
{
g.CopyFromScreen(SourcePoint, DestinationPoint, SelectionRectangle.Size);
if (showCursor)
{
Rectangle cursorBounds = new Rectangle(curPos, curSize);
Cursors.Default.Draw(g, cursorBounds);
}
}
using (MemoryStream ms = new MemoryStream())
{
Image img = (Image)bitmap;
img.Save(ms, ImageFormat.Jpeg);
Bitmap bm = new Bitmap(ms);
Clipboard.SetImage(bm);
if (!File.Exists(@"C:\Users\Cowboy\Desktop\Images\mstemp.jpg"))
{
img.Save(@"C:\Users\Cowboy\Desktop\Images\mstemp.jpg");
}
else
{
File.Delete(@"C:\Users\Cowboy\Desktop\Images\mstemp.jpg");
img.Save(@"C:\Users\Cowboy\Desktop\Images\mstemp.jpg");
}
ms.Dispose();
}
}
}
}
}

IronOCR and tesseract will not read the images.

var ocr = new AutoOcr();
var result = ocr.Read(@"C:\Users\Cowboy\Desktop\mstemp.jpg");
tbOCR.Text = result.GetType().ToString();

However, if IronOCR is able to read screenshot created by other applications like green shot, light shot, and print screen (a MS built in function), then the code I am using is in fact excluding some important raw data that would otherwise prevent the IronOCR functionality from doing its job. Nothing is wrong with IronOCR, its how the raw data stream/bitmap of my images are preventing the programs like IronOCR or Tesseract from reading my screenshots.

I think one of the following reasons might cause this issue:

1. The compression of the image might not be supported by the toolkit, or the image contains corrupted data.
After saving the image into JPEG format, try to open it using one of the commonly known viewers, such as MS Paint. If the image loads correctly, try to manually save it to PNG format and pass it to the IronOCR or the Tesseract toolkits and see if this fixes the OCR recognition issue.

2. Before saving the image data into MemoryStream, try to seek the stream to the beginning (origin) using the following method:

/////////////////////////////////////
ms.Seek(0, SeekOrigin.Begin);
// Save the image
// Pass it to the OCR toolkit
/////////////////////////////////////

3. If none of the above options fix the issue at your side, you can try using a different OCR programming toolkit that natively support both: loading different image formats (including RAW data) and text recognition (OCR). One of these good SDKs is LEADTOOLS SDK. The following code shows how you can pass your MemoryStream into the LEADTOOLS OCR engine, recognize it and then save the result into document format:

/////////////////////////////////////
IOcrEngine _ocrEngine = OcrEngineManager.CreateEngine(OcrEngineType.LEAD, false);
// Startup the engine
_ocrEngine.Startup(null, null, null, @"C:\LEADTOOLS 20\Bin\Common\OcrLEADRuntime");
using (IOcrDocument ocrDocument = _ocrEngine.DocumentManager.CreateDocument())
{
RasterImage _ocrimage = null;
using (RasterCodecs codecs = new RasterCodecs())
{
_ocrimage = codecs.Load(@"C:\Users\Cowboy\Desktop\mstemp.jpg");
// Or you can pass the MemoryStream directly after seeking it
// _ocrimage = codecs.Load(ms);
if (_ocrimage != null)
{
using (IOcrPage ocrPage = ocrDocument.Pages.AddPage(_ocrimage, null))
{
ocrPage.AutoZone(null);
ocrPage.Recognize(null);
Console.WriteLine(ocrPage.GetText(-1));
}
_ocrimage.Dispose();
}
}
/////////////////////////////////////

1 Reply

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

Maen Badwan8y ago
I think one of the following reasons might cause this issue:
1. The compression of the image might not be supported by the toolkit, or the image contains corrupted data.
After saving the image into JPEG format, try to open it using one of the commonly known viewers, such as MS Paint. If the image loads correctly, try to manually save it to PNG format and pass it to the IronOCR or the Tesseract toolkits and see if this fixes the OCR recognition issue.

2. Before saving the image data into MemoryStream, try to seek the stream to the beginning (origin) using the following method:
/////////////////////////////////////
ms.Seek(0, SeekOrigin.Begin);
// Save the image
// Pass it to the OCR toolkit
/////////////////////////////////////
3. If none of the above options fix the issue at your side, you can try using a different OCR programming toolkit that natively support both: loading different image formats (including RAW data) and text recognition (OCR). One of these good SDKs is LEADTOOLS SDK. The following code shows how you can pass your MemoryStream into the LEADTOOLS OCR engine, recognize it and then save the result into document format:
/////////////////////////////////////
IOcrEngine _ocrEngine = OcrEngineManager.CreateEngine(OcrEngineType.LEAD, false);
// Startup the engine
_ocrEngine.Startup(null, null, null, @"C:\LEADTOOLS 20\Bin\Common\OcrLEADRuntime");
using (IOcrDocument ocrDocument = _ocrEngine.DocumentManager.CreateDocument())
{
RasterImage _ocrimage = null;
using (RasterCodecs codecs = new RasterCodecs())
{
_ocrimage = codecs.Load(@"C:\Users\Cowboy\Desktop\mstemp.jpg");
// Or you can pass the MemoryStream directly after seeking it
// _ocrimage = codecs.Load(ms);
if (_ocrimage != null)
{
using (IOcrPage ocrPage = ocrDocument.Pages.AddPage(_ocrimage, null))
{
ocrPage.AutoZone(null);
ocrPage.Recognize(null);
Console.WriteLine(ocrPage.GetText(-1));
}
_ocrimage.Dispose();
}
}
/////////////////////////////////////