OCR: Recognize text in an image

In This Topic

Text recognition using OcrEngine class (TesseractOcr)

The OcrEngine.Recognize method must be called for the text recognition in the image. The method takes as input parameter the image with text and outputs the recognition results. The method performs the recognition in the working thread.

Here is C#/VB.NET code that shows how to recognize text in image:

string imageFilePath = @"D:\TestImage.png";
// create the OCR engine
using (Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr tesseractOcr = 
    new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr())
{
    // specify that OCR engine will recognize German text
    Vintasoft.Imaging.Ocr.OcrLanguage language = Vintasoft.Imaging.Ocr.OcrLanguage.German;
    // create the OCR engine settings
    Vintasoft.Imaging.Ocr.Tesseract.TesseractOcrSettings settings = 
        new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcrSettings(language);
    // initialize the OCR engine
    tesseractOcr.Init(settings);

    // load an image with text
    using (Vintasoft.Imaging.VintasoftImage image = new Vintasoft.Imaging.VintasoftImage(imageFilePath))
    {
        // specify the image, where text must be recognized
        tesseractOcr.SetImage(image);

        // recognize text in the image
        Vintasoft.Imaging.Ocr.Results.OcrPage ocrResult = tesseractOcr.Recognize();

        // get the recognized text
        string ocrResultAsText = ocrResult.GetText();

        string textFilePath = System.IO.Path.Combine(
            System.IO.Path.GetDirectoryName(imageFilePath),
            System.IO.Path.GetFileNameWithoutExtension(imageFilePath) + ".txt");
        // save the recognition results
        System.IO.File.WriteAllText(textFilePath, ocrResultAsText, System.Text.Encoding.UTF8);

        // clear the image
        tesseractOcr.ClearImage();
    }
    // shutdown the OCR engine
    tesseractOcr.Shutdown();
}

VB.NET

Dim imageFilePath As String = "D:\TestImage.png"
' create the OCR engine
Using tesseractOcr As New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr()
    ' specify that OCR engine will recognize German text
    Dim language As Vintasoft.Imaging.Ocr.OcrLanguage = Vintasoft.Imaging.Ocr.OcrLanguage.German
    ' create the OCR engine settings
    Dim settings As New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcrSettings(language)
    ' initialize the OCR engine
    tesseractOcr.Init(settings)

    ' load an image with text
    Using image As New Vintasoft.Imaging.VintasoftImage(imageFilePath)
        ' specify the image, where text must be recognized
        tesseractOcr.SetImage(image)

        ' recognize text in the image
        Dim ocrResult As Vintasoft.Imaging.Ocr.Results.OcrPage = tesseractOcr.Recognize()

        ' get the recognized text
        Dim ocrResultAsText As String = ocrResult.GetText()

        Dim textFilePath As String = System.IO.Path.Combine(System.IO.Path.GetDirectoryName(imageFilePath), System.IO.Path.GetFileNameWithoutExtension(imageFilePath) & ".txt")
        ' save the recognition results
        System.IO.File.WriteAllText(textFilePath, ocrResultAsText, System.Text.Encoding.UTF8)

        ' clear the image
        tesseractOcr.ClearImage()
    End Using
    ' shutdown the OCR engine
    tesseractOcr.Shutdown()
End Using

The OCR engine binarizes the image before the text recognition, if OcrEngine.Binarization property defines the image binarization command. The binarization will be disabled if value of the OcrEngine.Binarization property is set to null. The image binarization before text recognition in most cases improves the quality of text recognition for the document images. For mixed content images sometimes it is better do not perform image binarization before running OCR.

Text recognition using OcrEngineManager class

The OcrEngine.Recognize method performs the text recognition in the working thread. In a real situation, to speed up the recognition process, might be necessary to perform the text recognition in several threads. The same way OcrEngine.Recognize method performs the text recognition in a single language. In a real situation might be necessary to perform the text recognition in several languages at once. OcrEngineManager class is implemented to solve these problems.

OcrEngineManager class allows to:

Divide the image with text to several regions with text. For each region it is possible to:
- Define a bounding rectangle on image for the area where is necessary to recognize the text (RecognitionRegion.Rectangle)
- Specify the language of text (RecognitionRegion.Language)
- Specify the type of text region (RecognitionRegion.RegionType)
- Define the orthogonal rotation angle of text in the region (RecognitionRegion.TextRotation).
Recognize text in several regions
- By default the text recognition is performed in a single thread using only one instance of OCR engine. It is possible to recognize the text in several threads using several instances of OCR engine.

Text recognition in several threads

By default the text recognition is performed in a single thread using only one instance of OCR engine. It is possible to recognize the text in several threads using several instances of OCR engine.

Here is C#/VB.NET code that shows how to recognize text in several regions of the same image in two threads:

string imageFilePath = @"D:\TestImage.png";
// create the main OCR engine
using (Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr tesseractOcr = new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr())
{
    // create the additional OCR engine
    using (Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr additionalEngine = new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr())
    {
        // create the OCR engine manager
        Vintasoft.Imaging.Ocr.OcrEngineManager engineManager = 
            new Vintasoft.Imaging.Ocr.OcrEngineManager(tesseractOcr, additionalEngine);

        // load an image with text
        using (Vintasoft.Imaging.VintasoftImage image = new Vintasoft.Imaging.VintasoftImage(imageFilePath))
        {
            // create the OCR engine settings
            Vintasoft.Imaging.Ocr.OcrEngineSettings settings = new Vintasoft.Imaging.Ocr.OcrEngineSettings();
            // create the regions, where text must be searched
            Vintasoft.Imaging.Ocr.RecognitionRegion[] regions = new Vintasoft.Imaging.Ocr.RecognitionRegion[] {
                new Vintasoft.Imaging.Ocr.RecognitionRegion(
                    new Vintasoft.Imaging.RegionOfInterest(0,0, 319,80), Vintasoft.Imaging.Ocr.OcrLanguage.English),
                new Vintasoft.Imaging.Ocr.RecognitionRegion(
                    new Vintasoft.Imaging.RegionOfInterest(0, 330,319,85), Vintasoft.Imaging.Ocr.OcrLanguage.English, 180) };

            // recognize text in the image regions
            Vintasoft.Imaging.Ocr.Results.OcrPage ocrResult = engineManager.Recognize(image, settings, regions);

            // get the recognized text
            string ocrResultAsText = ocrResult.GetText();

            string textFilePath = System.IO.Path.Combine(
                System.IO.Path.GetDirectoryName(imageFilePath),
                System.IO.Path.GetFileNameWithoutExtension(imageFilePath) + ".txt");
            // save the recognition results
            System.IO.File.WriteAllText(textFilePath, ocrResultAsText, System.Text.Encoding.UTF8);
        }
    }
}

VB.NET

Dim imageFilePath As String = "D:\TestImage.png"
' create the main OCR engine
Using tesseractOcr As New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr()
    ' create the additional OCR engine
    Using additionalEngine As New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr()
        ' create the OCR engine manager
        Dim engineManager As New Vintasoft.Imaging.Ocr.OcrEngineManager(tesseractOcr, additionalEngine)

        ' load an image with text
        Using image As New Vintasoft.Imaging.VintasoftImage(imageFilePath)
            ' create the OCR engine settings
            Dim settings As New Vintasoft.Imaging.Ocr.OcrEngineSettings()
            ' create the regions, where text must be searched
            Dim regions As Vintasoft.Imaging.Ocr.RecognitionRegion() = New Vintasoft.Imaging.Ocr.RecognitionRegion() {New Vintasoft.Imaging.Ocr.RecognitionRegion(New Vintasoft.Imaging.RegionOfInterest(0, 0, 319, 80), Vintasoft.Imaging.Ocr.OcrLanguage.English), New Vintasoft.Imaging.Ocr.RecognitionRegion(New Vintasoft.Imaging.RegionOfInterest(0, 330, 319, 85), Vintasoft.Imaging.Ocr.OcrLanguage.English, 180)}

            ' recognize text in the image regions
            Dim ocrResult As Vintasoft.Imaging.Ocr.Results.OcrPage = engineManager.Recognize(image, settings, regions)

            ' get the recognized text
            Dim ocrResultAsText As String = ocrResult.GetText()

            Dim textFilePath As String = System.IO.Path.Combine(System.IO.Path.GetDirectoryName(imageFilePath), System.IO.Path.GetFileNameWithoutExtension(imageFilePath) & ".txt")
            ' save the recognition results
            System.IO.File.WriteAllText(textFilePath, ocrResultAsText, System.Text.Encoding.UTF8)
        End Using
    End Using
End Using

Here is C#/VB.NET code that shows how to recognize text in several images in three threads:

string imageFilePath = @"D:\TestImage.png";
// create the OCR engine
using (Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr tesseractOcr = 
    new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr())
{
    // create an array for additional OCR engines
    Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr[] additionalEngines = 
        new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr[2];
    try
    {
        // for each additional OCR engine
        for (int i = 0; i < additionalEngines.Length; i++)
            // create the additional OCR engine
            additionalEngines[i] = new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr();

        // create the OCR engine manager
        Vintasoft.Imaging.Ocr.OcrEngineManager engineManager = 
            new Vintasoft.Imaging.Ocr.OcrEngineManager(tesseractOcr, additionalEngines);
        // create an image collection
        using (Vintasoft.Imaging.ImageCollection images = 
            new Vintasoft.Imaging.ImageCollection())
        {
            // load the images from file
            images.Add(imageFilePath);

            // create the OCR engine settings
            Vintasoft.Imaging.Ocr.OcrEngineSettings settings = new Vintasoft.Imaging.Ocr.OcrEngineSettings();

            // recognize text in images
            Vintasoft.Imaging.Ocr.Results.OcrDocument ocrResults = engineManager.Recognize(images, settings);

            // create string with the recognition results

            System.Text.StringBuilder documentContent = new System.Text.StringBuilder();
            // for each recognized page
            for (int i = 0; i < ocrResults.Pages.Count; i++)
            {
                documentContent.AppendFormat("Page {0}", i + 1);
                documentContent.AppendLine();
                documentContent.AppendLine();

                documentContent.AppendLine(ocrResults.Pages[i].GetText());

                if (i != ocrResults.Pages.Count - 1)
                {
                    documentContent.AppendLine();
                    documentContent.AppendLine();
                    documentContent.AppendLine();
                }
            }

            string textFilePath = System.IO.Path.Combine(
                System.IO.Path.GetDirectoryName(imageFilePath),
                System.IO.Path.GetFileNameWithoutExtension(imageFilePath) + ".txt");
            // save the recognition results
            System.IO.File.WriteAllText(textFilePath, documentContent.ToString(), System.Text.Encoding.UTF8);

            // clear and dispose images
            images.ClearAndDisposeItems();
        }
    }
    finally
    {
        // for each additional OCR engine
        for (int i = 0; i < additionalEngines.Length; i++)
        {
            if (additionalEngines[i] != null)
                // dispose the additional OCR engine
                additionalEngines[i].Dispose();
        }
    }
}

VB.NET

Dim imageFilePath As String = "D:\TestImage.png"
' create the OCR engine
Using tesseractOcr As New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr()
    ' create an array for additional OCR engines
    Dim additionalEngines As Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr() = New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr(1) {}
    Try
        ' for each additional OCR engine
        For i As Integer = 0 To additionalEngines.Length - 1
            ' create the additional OCR engine
            additionalEngines(i) = New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr()
        Next

        ' create the OCR engine manager
        Dim engineManager As New Vintasoft.Imaging.Ocr.OcrEngineManager(tesseractOcr, additionalEngines)
        ' create an image collection
        Using images As New Vintasoft.Imaging.ImageCollection()
            ' load the images from file
            images.Add(imageFilePath)

            ' create the OCR engine settings
            Dim settings As New Vintasoft.Imaging.Ocr.OcrEngineSettings()

            ' recognize text in images
            Dim ocrResults As Vintasoft.Imaging.Ocr.Results.OcrDocument = engineManager.Recognize(images, settings)

            ' create string with the recognition results

            Dim documentContent As New System.Text.StringBuilder()
            ' for each recognized page
            For i As Integer = 0 To ocrResults.Pages.Count - 1
                documentContent.AppendFormat("Page {0}", i + 1)
                documentContent.AppendLine()
                documentContent.AppendLine()

                documentContent.AppendLine(ocrResults.Pages(i).GetText())

                If i <> ocrResults.Pages.Count - 1 Then
                    documentContent.AppendLine()
                    documentContent.AppendLine()
                    documentContent.AppendLine()
                End If
            Next

            Dim textFilePath As String = System.IO.Path.Combine(System.IO.Path.GetDirectoryName(imageFilePath), System.IO.Path.GetFileNameWithoutExtension(imageFilePath) & ".txt")
            ' save the recognition results
            System.IO.File.WriteAllText(textFilePath, documentContent.ToString(), System.Text.Encoding.UTF8)

            ' clear and dispose images
            images.ClearAndDisposeItems()
        End Using
    Finally
        ' for each additional OCR engine
        For i As Integer = 0 To additionalEngines.Length - 1
            If additionalEngines(i) IsNot Nothing Then
                ' dispose the additional OCR engine
                additionalEngines(i).Dispose()
            End If
        Next
    End Try
End Using

Text recognition in very large image

Tesseract OCR creates a copy of image data before text recognition. This may lead to a huge use of memory or even memory overflow in case the image is very large. OcrEngineManager class allows to divide the very large image into several not so large images, recognize the text in the resulting images and unite the OCR results into one total result.

Here is C#/VB.NET code that shows how to recognize text in a very large image (20000x30000, 24-bpp, 600 dpi):

string imageFilePath = @"D:\TestImage.png";
// create the OCR engine
using (Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr tesseractOcr = new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr())
{
    // create an array for additional OCR engines
    Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr[] additionalEngines = new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr[3];
    try
    {
        // for each additional OCR engine
        for (int i = 0; i < additionalEngines.Length; i++)
            // create the additional OCR engine
            additionalEngines[i] = new Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr();

        // create the OCR engine manager
        Vintasoft.Imaging.Ocr.OcrEngineManager engineManager = new Vintasoft.Imaging.Ocr.OcrEngineManager(tesseractOcr, additionalEngines);

        // load an image from file
        using (Vintasoft.Imaging.VintasoftImage image = new Vintasoft.Imaging.VintasoftImage(imageFilePath))
        {
            // create the recognition region splitting settings
            Vintasoft.Imaging.Ocr.OcrRecognitionRegionSplittingSettings splittingSettings =
                new Vintasoft.Imaging.Ocr.OcrRecognitionRegionSplittingSettings(5000, 5000, 2000);
            // specify that the recognition regions must be splitted into sub regions
            engineManager.RecognitionRegionSplittingSettings = splittingSettings;

            // create the OCR engine settings
            Vintasoft.Imaging.Ocr.OcrEngineSettings settings = new Vintasoft.Imaging.Ocr.OcrEngineSettings();

            // recognize text in image
            Vintasoft.Imaging.Ocr.Results.OcrPage ocrResult = engineManager.Recognize(image, settings);

            // get the recognized text
            string ocrResultAsText = ocrResult.GetText();

            string textFilePath = System.IO.Path.Combine(
                System.IO.Path.GetDirectoryName(imageFilePath),
                System.IO.Path.GetFileNameWithoutExtension(imageFilePath) + ".txt");
            // save the recognized text in file
            System.IO.File.WriteAllText(textFilePath, ocrResultAsText, System.Text.Encoding.UTF8);
        }
    }
    finally
    {
        // for each additional OCR engine
        for (int i = 0; i < additionalEngines.Length; i++)
        {
            if (additionalEngines[i] != null)
                // dispose the additional OCR engine
                additionalEngines[i].Dispose();
        }
    }
}

VB.NET

Dim imageFilePath As String = "D:\TestImage.png"
' create the OCR engine
Using tesseractOcr As New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr()
    ' create an array for additional OCR engines
    Dim additionalEngines As Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr() = New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr(2) {}
    Try
        ' for each additional OCR engine
        For i As Integer = 0 To additionalEngines.Length - 1
            ' create the additional OCR engine
            additionalEngines(i) = New Vintasoft.Imaging.Ocr.Tesseract.TesseractOcr()
        Next

        ' create the OCR engine manager
        Dim engineManager As New Vintasoft.Imaging.Ocr.OcrEngineManager(tesseractOcr, additionalEngines)

        ' load an image from file
        Using image As New Vintasoft.Imaging.VintasoftImage(imageFilePath)
            ' create the recognition region splitting settings
            Dim splittingSettings As New Vintasoft.Imaging.Ocr.OcrRecognitionRegionSplittingSettings(5000, 5000, 2000)
            ' specify that the recognition regions must be splitted into sub regions
            engineManager.RecognitionRegionSplittingSettings = splittingSettings

            ' create the OCR engine settings
            Dim settings As New Vintasoft.Imaging.Ocr.OcrEngineSettings()

            ' recognize text in image
            Dim ocrResult As Vintasoft.Imaging.Ocr.Results.OcrPage = engineManager.Recognize(image, settings)

            ' get the recognized text
            Dim ocrResultAsText As String = ocrResult.GetText()

            Dim textFilePath As String = System.IO.Path.Combine(System.IO.Path.GetDirectoryName(imageFilePath), System.IO.Path.GetFileNameWithoutExtension(imageFilePath) & ".txt")
            ' save the recognized text in file
            System.IO.File.WriteAllText(textFilePath, ocrResultAsText, System.Text.Encoding.UTF8)
        End Using
    Finally
        ' for each additional OCR engine
        For i As Integer = 0 To additionalEngines.Length - 1
            If additionalEngines(i) IsNot Nothing Then
                ' dispose the additional OCR engine
                additionalEngines(i).Dispose()
            End If
        Next
    End Try
End Using

Send Feedback