C# 开源 OCR（开发者列表）

Tesseract

The most renowned open-source OCR engine is called Tesseract, and it was initially created by Hewlett-Packard. Since 2006, Google has been supporting this free software project, which is released under the Apache license.

One of the most accurate open-source and free systems available is the Tesseract OCR engine. Tesseract now supports 116 languages with its most recent stable version, 4.1.1, which is based on LSTM.

Tesseract requires support from a separate GUI (graphical user interface) when running from a command-line interface because it does not have its own built-in interface. It can learn new information using its neural networks and has an advanced image preprocessing pipeline. The most effective technique to add OCR capabilities to your .NET application is the Tesseract .NET SDK, which is one of the best solutions for providing text recognition capabilities. Even though Tesseract is undoubtedly the best OCR library currently on the market.

GOCR

The GNU Public License was used to create the OCR (Optical Character Recognition) program known as GOCR. It transforms text files back into scanned images of documents. After starting the program and managing the development team on SF, Joerg Schulenburg continues to handle the package at a (very) low time base today.

Since GOCR can be used with several front-ends, it is relatively simple to port it to other operating systems, network applications, and architectures. It can read a wide range of picture file types, and until 2010, its quality consistently improved.

According to GOCR, it can handle single-column sans-serif fonts with a height of 20–60 pixels. It reports difficulties with text written in alphabets other than Latin, serif fonts, overlapping letters, various typefaces, noisy photos, and excessive angles of skew. GOCR is also capable of translating barcodes.

CuneiForm

CuneiForm, a free and open-source technology, is now also known as "Cognitive OpenOCR." It has built-in output and a database. It covers 23 distinct languages and also performs tasks such as text format scanning, document layout analysis, and identification.

Cognitive Technologies developed the licenses for OpenOCR, which are freeware and BSD. While it supports cross-platform use, Linux users are not provided with a graphical interface.

To simplify character recognition work in any Dot NET Framework 2.0 or later applications, the wrapper library Puma Dot NET is used. It runs a dictionary check while processing data to enhance the quality of recognition.

CuneiForm is a technology designed to automatically or semi-automatically convert electronic copies of paper documents and image files into an editable form without affecting the structure and original document fonts. The system consists of two parts for processing electronic documents in batches and one document at a time. Furthermore, the system supports a combination of Russian and English. Only the branch created by Andrei Borovsky in 2009 supports the recognition of other hybrid languages. Teaching the system to recognize other languages is challenging since each language is associated with a dat-file, the structure and creation process of which are not disclosed by the developers.

Kraken

Kraken was developed to address the issues with Ocropus without impacting its other features. It utilizes its CLSTM neural network library and leverages the valuable experience gained from prior projects with fresh data. It requires the use of certain external libraries to function effectively across different platforms. With the help of the stored information, it can make more accurate predictions regarding potential data validation problems. Furthermore, its working methodology facilitates the easy deployment and training of new models.

IronOCR

In contrast to the standard Tesseract library, IronOCR expands Tesseract and provides a native C# OCR library with higher accuracy, improved performance, and enhanced stability. IronOCR can be used in .NET programs and websites to extract text from PDFs and images. It supports a wide range of foreign languages and can generate plain text or structured data output. It is capable of scanning barcodes and images with embedded text. The library can be utilized in applications developed in .NET for the console, web, MVC, and desktop. The development team offers direct assistance with the licensing process for commercial deployments. IronOCR is compatible with the latest versions of Visual Studio.

using IronOcr;

// Instantiate an IronTesseract object to utilize its OCR capabilities
var Ocr = new IronTesseract();

// Set the language to English for better accuracy
Ocr.Language = OcrLanguage.EnglishBest;

// Optionally specify the Tesseract version to ensure compatibility
Ocr.Configuration.TesseractVersion = TesseractVersion.Tesseract5;

// Create an OcrInput object to add images for OCR processing
using (var Input = new OcrInput())
{
    // Add the image to be processed; specify the image's path
    Input.AddImage(@"Demo.png");

    // Perform the OCR and store the result
    var Result = Ocr.Read(Input);

    // Output the extracted text to the console
    Console.WriteLine(Result.Text);

    // Pause the console to keep it open
    Console.ReadKey();
}

using IronOcr;

// Instantiate an IronTesseract object to utilize its OCR capabilities
var Ocr = new IronTesseract();

// Set the language to English for better accuracy
Ocr.Language = OcrLanguage.EnglishBest;

// Optionally specify the Tesseract version to ensure compatibility
Ocr.Configuration.TesseractVersion = TesseractVersion.Tesseract5;

// Create an OcrInput object to add images for OCR processing
using (var Input = new OcrInput())
{
    // Add the image to be processed; specify the image's path
    Input.AddImage(@"Demo.png");

    // Perform the OCR and store the result
    var Result = Ocr.Read(Input);

    // Output the extracted text to the console
    Console.WriteLine(Result.Text);

    // Pause the console to keep it open
    Console.ReadKey();
}

Imports IronOcr

' Instantiate an IronTesseract object to utilize its OCR capabilities
Private Ocr = New IronTesseract()

' Set the language to English for better accuracy
Ocr.Language = OcrLanguage.EnglishBest

' Optionally specify the Tesseract version to ensure compatibility
Ocr.Configuration.TesseractVersion = TesseractVersion.Tesseract5

' Create an OcrInput object to add images for OCR processing
Using Input = New OcrInput()
	' Add the image to be processed; specify the image's path
	Input.AddImage("Demo.png")

	' Perform the OCR and store the result
	Dim Result = Ocr.Read(Input)

	' Output the extracted text to the console
	Console.WriteLine(Result.Text)

	' Pause the console to keep it open
	Console.ReadKey()
End Using

Conclusion

OCR open-source tools allow us to build our own programs using their source code. However, some tools do not have an official library or dedicated team to provide support in case of coding issues. Tesseract's documentation also lacks sample code or tutorials for common use scenarios, making it challenging for beginners to understand the code and libraries.

IronOCR supports various .NET projects such as .NET Framework Standard 2, .NET Framework 4.5, and .NET Core 2, 3, and 5. It also works with newer technologies like Mono, Xamarin, and Azure. By leveraging IronOCR technologies, we can enhance Tesseract's results and correct inaccurately scanned documents or images. The complex Tesseract dictionary system is managed through the NuGet Package. We utilize the Iron OCR Library to develop an OCR tool.

With IronOCR, we can use the program without any additional configuration, and it supports PDF files, multi-frame TIFF, and all common image formats. It also offers barcode recognition capabilities, allowing us to extract barcode data and read barcode values from images. IronOCR provides a cost-effective development edition with a free trial, and the lifetime license is included in the IronOCR bundle at no extra cost. The IronOCR bundle provides coverage for multiple platforms with a single payment. For more information on IronOCR's pricing, please refer to this page.