C#でOCRを使用した文字認識の作成方法

Improved Tesseract OCR Engine

IronOCR elevates the capabilities of the widely used Tesseract OCR engine by enhancing both accuracy and speed. It serves as a robust solution for extracting text from various sources, including images, PDFs, and diverse document formats.

Wide Language Coverage

With support for over 125 languages, IronOCR is adept at handling multilingual requirements, making it an ideal choice for applications demanding linguistic versatility.

Versatile Output Choices

Extracted text can be conveniently outputted as plain text or structured data for seamless integration into further processing pipelines. Additionally, IronOCR facilitates the creation of searchable PDFs directly from image inputs.

Cross-Platform Adaptability

Engineered for compatibility with C#, F#, and VB.NET, IronOCR seamlessly operates across various .NET environments including versions 8, 7, 6, Core, Standard, and Framework.

Leveraging Tesseract 5

IronOCR harnesses the power of Tesseract 5, finely tailored for optimal performance within the .NET ecosystem.

Zone-Based OCR Capability

With IronOCR, users can precisely define specific zones within documents, enabling targeted OCR processing. This feature enhances accuracy and efficiency by focusing processing power where it's needed most.

Image Preprocessing Tools

The library offers a suite of image preprocessing functionalities such as de-skewing and noise reduction. These tools ensure superior results even when dealing with imperfect source images, ultimately enhancing the overall OCR experience.

Now, we will develop a demo application that utilizes IronOCR to read text from images.

Code Explanation

1. IronTesseract Instance: We start by creating an instance of IronTesseract to perform OCR operations.
2. Loading Image: We load the sample image into the OcrInput object.
3. Reading Text: The text in the image is read, and the result is printed to the console.

Output

Language Selection

You can specify the language for OCR using the Language property. For instance, to set the language to English, use:

IronTesseract ocr = new IronTesseract();
ocr.Language = OcrLanguage.English;

IronTesseract ocr = new IronTesseract();
ocr.Language = OcrLanguage.English;

Dim ocr As New IronTesseract()
ocr.Language = OcrLanguage.English

Page Segmentation Mode

The PageSegmentationMode determines how Tesseract segments the input image. Options include AutoOsd, SingleBlock, SingleLine, and more. For example:

ocr.Configuration.PageSegmentationMode = TesseractPageSegmentationMode.AutoOsd;

ocr.Configuration.PageSegmentationMode = TesseractPageSegmentationMode.AutoOsd;

ocr.Configuration.PageSegmentationMode = TesseractPageSegmentationMode.AutoOsd

Custom Tesseract Variables

You can fine-tune Tesseract by setting specific variables. For instance, to disable parallelization:

ocr.Configuration.TesseractVariables["tessedit_parallelize"] = false;

ocr.Configuration.TesseractVariables["tessedit_parallelize"] = false;

ocr.Configuration.TesseractVariables("tessedit_parallelize") = False

Whitelisting and Blacklisting Characters

Use WhiteListCharacters and BlackListCharacters to control which characters Tesseract recognizes. For example:

ocr.Configuration.WhiteListCharacters = "ABCDEFGHIJKLMNOPQRSTUVWXYZ";
ocr.Configuration.BlackListCharacters = "`ë|^";

ocr.Configuration.WhiteListCharacters = "ABCDEFGHIJKLMNOPQRSTUVWXYZ";
ocr.Configuration.BlackListCharacters = "`ë|^";

ocr.Configuration.WhiteListCharacters = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
ocr.Configuration.BlackListCharacters = "`ë|^"

Additional Configuration Variables

Explore other Tesseract configuration variables to customize behavior according to your needs. For instance:

ocr.Configuration.TesseractVariables["classify_num_cp_levels"] = 3;
ocr.Configuration.TesseractVariables["textord_debug_tabfind"] = 0;
// ... (more variables)

ocr.Configuration.TesseractVariables["classify_num_cp_levels"] = 3;
ocr.Configuration.TesseractVariables["textord_debug_tabfind"] = 0;
// ... (more variables)

ocr.Configuration.TesseractVariables("classify_num_cp_levels") = 3
ocr.Configuration.TesseractVariables("textord_debug_tabfind") = 0
' ... (more variables)

Now let us try to decode the same image using advanced settings

using IronOcr;

public class Program
{
    public static void Main()
    {
        Console.WriteLine("Decoding using advanced features");
        var ocrTesseract = new IronTesseract() // Create instance
        {
            Language = OcrLanguage.EnglishBest, // Configure best English language
            Configuration = new TesseractConfiguration()
            {
                ReadBarCodes = false, // Disable reading barcodes
                BlackListCharacters = "`ë|^", // Blacklisted characters
                WhiteListCharacters = null, // No whitelist, allow all
                PageSegmentationMode = TesseractPageSegmentationMode.AutoOsd,
                TesseractVariables = null, // No custom variables used
            },
            MultiThreaded = false,
        };

        using var ocrInput = new OcrInput(); // Create a disposible ocr input object
        ocrInput.AddImage(@"sample1.png"); // Load the sample image
        var ocrResult = ocrTesseract.Read(ocrInput); // Read the text from the image
        Console.WriteLine(ocrResult.Text); // Output the text
    }
}

using IronOcr;

public class Program
{
    public static void Main()
    {
        Console.WriteLine("Decoding using advanced features");
        var ocrTesseract = new IronTesseract() // Create instance
        {
            Language = OcrLanguage.EnglishBest, // Configure best English language
            Configuration = new TesseractConfiguration()
            {
                ReadBarCodes = false, // Disable reading barcodes
                BlackListCharacters = "`ë|^", // Blacklisted characters
                WhiteListCharacters = null, // No whitelist, allow all
                PageSegmentationMode = TesseractPageSegmentationMode.AutoOsd,
                TesseractVariables = null, // No custom variables used
            },
            MultiThreaded = false,
        };

        using var ocrInput = new OcrInput(); // Create a disposible ocr input object
        ocrInput.AddImage(@"sample1.png"); // Load the sample image
        var ocrResult = ocrTesseract.Read(ocrInput); // Read the text from the image
        Console.WriteLine(ocrResult.Text); // Output the text
    }
}

Imports IronOcr

Public Class Program
	Public Shared Sub Main()
		Console.WriteLine("Decoding using advanced features")
		Dim ocrTesseract = New IronTesseract() With {
			.Language = OcrLanguage.EnglishBest,
			.Configuration = New TesseractConfiguration() With {
				.ReadBarCodes = False,
				.BlackListCharacters = "`ë|^",
				.WhiteListCharacters = Nothing,
				.PageSegmentationMode = TesseractPageSegmentationMode.AutoOsd,
				.TesseractVariables = Nothing
			},
			.MultiThreaded = False
		}

		Dim ocrInput As New OcrInput() ' Create a disposible ocr input object
		ocrInput.AddImage("sample1.png") ' Load the sample image
		Dim ocrResult = ocrTesseract.Read(ocrInput) ' Read the text from the image
		Console.WriteLine(ocrResult.Text) ' Output the text
	End Sub
End Class

Code Explanation

1. IronOCR Configuration: An instance of IronTesseract (the main IronOCR class) is created and assigned to the variable ocrTesseract.

   Configuration settings are applied to ocrTesseract:

   • Language: Specifies the language for OCR (in this case, English).
   • Configuration: A TesseractConfiguration object that allows further customization:
     • ReadBarCodes: Disables reading barcodes.
     • BlackListCharacters: Specifies characters to blacklist (characters not to recognize).
     • WhiteListCharacters: No whitelist specified, allowing all characters.
     • PageSegmentationMode: Sets the page segmentation mode to "AutoOsd."
     • TesseractVariables: No custom variables were used.
   • MultiThreaded: Disables multithreading.
2. OCR Input and Image Loading: A using block creates a disposable ocrInput object of type OcrInput. The image file "sample1.png" is added to ocrInput.
3. Text Extraction: The Read method is called on ocrTesseract, passing in ocrInput. The result is stored in the ocrResult variable.
4. Output: The extracted text is printed to the console using Console.WriteLine(ocrResult.Text).

Output

Step 5: Performance Tuning of IronOCR Read Operation.

When working with IronOCR, you have access to various image filters that can help preprocess images before performing OCR. These filters optimize the image quality, enhance visibility, and reduce noise or artifacts. They help to improve the performance of the OCR operation.

1. Rotate:

   The Rotate filter allows you to rotate images by a specified number of degrees clockwise. For anti-clockwise rotation, use negative numbers.

2. Deskew:

   The Deskew filter corrects image skew, ensuring that the text is upright and orthogonal. This is particularly useful for OCR because Tesseract performs best with properly oriented scans.

3. Scale:

   The Scale filter proportionally scales OCR input pages.

4. Binarize:

   The Binarize filter converts every pixel to either black or white, with no middle ground. It can improve OCR performance in cases of very low contrast between text and background.

5. ToGrayScale:

   The ToGrayScale filter converts every pixel to a shade of grayscale. While unlikely to significantly improve OCR accuracy, it may enhance speed.

6. Invert:

   The Invert filter reverses colors—white becomes black, and black becomes white.

7. ReplaceColor:

   The ReplaceColor filter replaces a specific color within an image with another color, considering a certain threshold.

8. Contrast:

   The Contrast filter automatically increases contrast. It often improves OCR speed and accuracy in low-contrast scans.

9. Dilate and Erode:

   These advanced morphology filters manipulate object boundaries in an image.

   • Dilate adds pixels to object boundaries.
   • Erode removes pixels from object boundaries.

10. Sharpen:

    The Sharpen filter sharpens blurred OCR documents and flattens alpha channels to white.

11. DeNoise:

    The DeNoise filter removes digital noise. Use it where noise is expected.

12. DeepCleanBackgroundNoise:

    This heavy background noise removal filter should be used only when extreme document background noise is known. It may reduce OCR accuracy for clean documents and is CPU-intensive.

13. EnhanceResolution:

    The EnhanceResolution filter enhances the resolution of low-quality images. It’s not often needed due to automatic resolution handling.

Here’s an example of how to apply filters using IronOCR in C#:

var ocr = new IronTesseract();
var input = new OcrInput();
input.LoadImage("sample.png");
input.Deskew();
var result = ocr.Read(input);
Console.WriteLine(result.Text);

var ocr = new IronTesseract();
var input = new OcrInput();
input.LoadImage("sample.png");
input.Deskew();
var result = ocr.Read(input);
Console.WriteLine(result.Text);

Dim ocr = New IronTesseract()
Dim input = New OcrInput()
input.LoadImage("sample.png")
input.Deskew()
Dim result = ocr.Read(input)
Console.WriteLine(result.Text)

Common OCR Applications

1. Document Digitization: OCR is widely used to convert scanned paper documents, such as invoices, receipts, forms, and contracts, into digital formats. This digitization process streamlines document storage, retrieval, and management, reducing paper clutter and improving efficiency.
2. Data Extraction: OCR enables the extraction of text and data from scanned documents, images, and PDFs. This extracted data can be used for automated data entry, content analysis, indexing, and integration into databases or business systems.
3. Text Recognition in Images: OCR technology allows extracting text from printed documents and images for indexing and search purposes. This capability is utilized in various applications, including augmented reality, image-based search engines, and translation services.
4. Automatic License Plate Recognition (ALPR): ALPR systems utilize OCR to read license plate numbers from images or video streams captured by cameras installed in traffic surveillance, parking management, toll collection, and law enforcement applications.
5. Accessibility Solutions: OCR plays a crucial role in creating accessible content for individuals with visual impairments. By converting text from images or documents into speech or braille, OCR helps make information accessible to people with disabilities.
6. Identity Verification: OCR technology is employed in identity verification processes, such as scanning and processing identity documents like passports, driver's licenses, and IDs. It assists in verifying the authenticity of documents and extracting relevant information for identity verification purposes.
7. Banking and Finance: OCR is used in banking and finance for tasks such as reading checks, processing invoices, converting an existing PDF document, extracting data from financial statements, and automating document-based workflows to enhance accuracy and efficiency in financial operations.
8. Automated Translation: OCR technology is integrated into translation tools and language learning apps to convert printed text from one language to another. Users can capture text with their devices, and OCR assists in translating it into the desired language in real time.
9. Archival and Historical Document Preservation: OCR is utilized in digitizing archival materials and historical documents, preserving them in digital formats for future access, research, and analysis while ensuring the preservation of valuable cultural heritage.