1. IronOCR
  2. How-Tos
  3. C# PDF OCR

How to Build an Azure OCR Service using IronOCR

Iron Software has created an OCR (Optical Character Recognition) library that takes the interoperability issues out of Azure OCR integration. Working with OCR libraries on Azure has always been a bit of a pain for developers. The solution for this and many other OCR headaches is IronOCR.

IronOCR features for Microsoft Azure

IronOCR includes the following features for building and OCR Service on Microsoft Azure:

  • Turns PDFs into searchable documents so that it is easy to extract text
  • Turns images into searchable documents by extracting text from images
  • Reads barcodes as well as QR codes
  • Exceptional accuracy
  • Runs locally and requires no SaaS (Software as a Service) which is a software distribution model where a cloud provider, such as Microsoft Azure, hosts various applications and makes these applications available to end users.
  • Lightning-fast speed

Let’s have a look at how the best OCR engine, Iron Software’s IronOCR, makes it easier for developers to extract text from any input document.

Let’s get started with our Azure OCR Service

In order to get started with the sample, we need to install IronOCR first.

  1. Create a new Console application with C#
  2. Install IronOCR via NuGet either by entering: Install-Package IronOcr or by selecting Manage NuGet packages and search for IronOCR. This is shown below
  3. Edit your Program.cs file to look like the following:
    • We import the IronOcr namespace to make use of its ocr capabilities to read and extract the contents of the PDF file.
    • We create a new IronTesseract object, so that we can extract text from an image.
using IronOcr;
using System;

namespace IronOCR_Ex
{
    class Program
    {
        static void Main(string [] args)
        {
            var ocr = new IronTesseract();
            using (var Input = new OcrInput("..\\Images\\Purgatory.PNG"))
            {
                var result = ocr.Read(Input); //Read PNG image File
                Console.WriteLine(result.Text); //Write Output to PDF document
                Console.ReadLine();
            }
        }
    }
}
using IronOcr;
using System;

namespace IronOCR_Ex
{
    class Program
    {
        static void Main(string [] args)
        {
            var ocr = new IronTesseract();
            using (var Input = new OcrInput("..\\Images\\Purgatory.PNG"))
            {
                var result = ocr.Read(Input); //Read PNG image File
                Console.WriteLine(result.Text); //Write Output to PDF document
                Console.ReadLine();
            }
        }
    }
}
Imports IronOcr
Imports System

Namespace IronOCR_Ex
	Friend Class Program
		Shared Sub Main(ByVal args() As String)
			Dim ocr = New IronTesseract()
			Using Input = New OcrInput("..\Images\Purgatory.PNG")
				Dim result = ocr.Read(Input) 'Read PNG image File
				Console.WriteLine(result.Text) 'Write Output to PDF document
				Console.ReadLine()
			End Using
		End Sub
	End Class
End Namespace
VB   C#
  1. Next, we open an image named Purgatory.PNG. This image forms part of the Devine Comedy by Dante - one of my favorite books. The picture looks like the next image.

The text to be extracted with the optical character reading capabilities of IronOCR Figure 2 - The text to be extracted with the optical character reading capabilities of IronOCR

  1. The output after the above text has been extracted from the above input image text.

Extracted text Figure 3 - Extracted text

  1. Let’s do the same with a PDF document. The PDF document contains the same text to extract as Figure.

The only difference is that we will be a PDF document instead of an image. Enter the following code:

 var Ocr = new IronTesseract();
            using (var input = new OcrInput())
            {
                input.Title = "Divine Comedy - Purgatory"; //Give title to input document 
                //Supply optional password and name of document
                input.AddPdf("..\\Documents\\Purgatorio.pdf", "dante");
                var Result = Ocr.Read(input); //Read the input file

                Result.SaveAsSearchablePdf("SearchablePDFDocument.pdf"); 
            }
 var Ocr = new IronTesseract();
            using (var input = new OcrInput())
            {
                input.Title = "Divine Comedy - Purgatory"; //Give title to input document 
                //Supply optional password and name of document
                input.AddPdf("..\\Documents\\Purgatorio.pdf", "dante");
                var Result = Ocr.Read(input); //Read the input file

                Result.SaveAsSearchablePdf("SearchablePDFDocument.pdf"); 
            }
Dim Ocr = New IronTesseract()
			Using input = New OcrInput()
				input.Title = "Divine Comedy - Purgatory" 'Give title to input document
				'Supply optional password and name of document
				input.AddPdf("..\Documents\Purgatorio.pdf", "dante")
				Dim Result = Ocr.Read(input) 'Read the input file

				Result.SaveAsSearchablePdf("SearchablePDFDocument.pdf")
			End Using
VB   C#

Almost the same as the previous code that extracts text from an image.

Here we make use of the OcrInput method to read the current PDF document, in this case: Purgatorio.pdf. If there is metadata in the PDF file such as a title or a password, we can also feed it in.

The result gets saved as a PDF document in which we can search for text.

Note, if the PDF file is too big, an exception may be thrown.

  1. Enough on Windows applications; let’s have a look at how we can use ocr with Microsoft Azure.

The beauty of IronOCR is that it works very well with Microsoft Azure as an Azure Function in a microservice architecture. Here is a very quick example on what a Microsoft Azure Function that works with IronOCR would look like. This Microsoft Azure function extracts text from images.

public static class OCRFunction
{
    public static HttpClient hcClient = new HttpClient();

    [FunctionName("IronOCRFunction_EX")]
    public static async Task<IActionResult> Run([HttpTrigger] HttpRequest hrRequest, ExecutionContext ecContext)
    {
        var URI = hrRequest.Query ["image"];
        var saStream = await hcClient.GetStreamAsync(URI);

        var ocr = new IronTesseract();
        using (var inputOCR = new OcrInput(saStream))
        {
            var outputOCR = ocr.Read(inputOCR);
            return new OkObjectResult(outputOCR.Text);
        }
    }
} 
public static class OCRFunction
{
    public static HttpClient hcClient = new HttpClient();

    [FunctionName("IronOCRFunction_EX")]
    public static async Task<IActionResult> Run([HttpTrigger] HttpRequest hrRequest, ExecutionContext ecContext)
    {
        var URI = hrRequest.Query ["image"];
        var saStream = await hcClient.GetStreamAsync(URI);

        var ocr = new IronTesseract();
        using (var inputOCR = new OcrInput(saStream))
        {
            var outputOCR = ocr.Read(inputOCR);
            return new OkObjectResult(outputOCR.Text);
        }
    }
} 
Public Module OCRFunction
	Public hcClient As New HttpClient()

	<FunctionName("IronOCRFunction_EX")>
	Public Async Function Run(<HttpTrigger> ByVal hrRequest As HttpRequest, ByVal ecContext As ExecutionContext) As Task(Of IActionResult)
		Dim URI = hrRequest.Query ("image")
		Dim saStream = Await hcClient.GetStreamAsync(URI)

		Dim ocr = New IronTesseract()
		Using inputOCR = New OcrInput(saStream)
			Dim outputOCR = ocr.Read(inputOCR)
			Return New OkObjectResult(outputOCR.Text)
		End Using
	End Function
End Module
VB   C#

This feeds the image received by the function directly to the ocr engine to be output as extracted text.

A quick recap on Microsoft Azure.

According to Microsoft: Microsoft Azure Microservices are an architectural approach to building applications where each core function, or service, is built and deployed independently. Microservice architecture is distributed and loosely coupled, so one component’s failure won’t break the whole app. independent components work together and communicate with well-defined API contracts. Build microservice applications to meet rapidly changing business needs and bring new functionalities to market faster.

A few more features of IronOCR with .NET or Microsoft Azure include the following:

The ability to perform ocr on almost any file, image, or PDF.

  • Lightning-fast speed in processing ocr input
  • Exceptional accuracy
  • Reads barcodes and QR codes
  • Runs locally, with no SaaS required
  • Can turn PDFs and images into searchable documents
  • Excellent Alternative to Azure OCR from Microsoft Cognitive Services

Image Filters to improve OCR performance

  • OcrInput.Rotate - Rotates images by several degrees clockwise. For anti-clockwise, use negative numbers.
  • OcrInput.Binarize() - This image filter turns every pixel black or white with no middle ground. This improves OCR performance.
  • OcrInput.ToGrayScale() - This image filter turns every pixel into a shade of grayscale. This improves OCR speed
  • OcrInput.Contrast() - Increases contrast automatically. This filter improves ocr speed and accuracy in low contrast scans.
  • OcrInput.DeNoise() - Removes digital noise. This filter should only be used where noise is expected in input documents.
  • OcrInput.Invert() - Inverts every color.
  • OcrInput.Dilate() - Dilation adds pixels to the boundaries of any object in an image.
  • OcrInput.Erode() - Erosion removes pixels on object boundaries.
  • OcrInput.Deskew() - Rotates an image so it is the right way up and orthogonal. This is very useful for OCR because Tesseract tolerance for skewed scans can be as low as 5 degrees.
  • OcrInput.DeepCleanBackgroundNoise() - Heavy background noise removal.
  • OcrInput.EnhanceResolution - Enhances the resolution of a low-quality image.

Speed performance

An example follows:

    var Ocr = new IronTesseract();
    Ocr.Configuration.BlackListCharacters = "~`$#^*_}{][|\\";
    Ocr.Configuration.PageSegmentationMode = TesseractPageSegmentationMode.Auto;
    Ocr.Configuration.TesseractVersion = TesseractVersion.Tesseract5;
    Ocr.Configuration.EngineMode = TesseractEngineMode.LstmOnly;
    Ocr.Language = OcrLanguage.EnglishFast;
    using (var Input = new OcrInput("..\\Images\\Purgatory.PNG"))
    {
        var Result = Ocr.Read(Input);
        Console.WriteLine(Result.Text);
    }
    var Ocr = new IronTesseract();
    Ocr.Configuration.BlackListCharacters = "~`$#^*_}{][|\\";
    Ocr.Configuration.PageSegmentationMode = TesseractPageSegmentationMode.Auto;
    Ocr.Configuration.TesseractVersion = TesseractVersion.Tesseract5;
    Ocr.Configuration.EngineMode = TesseractEngineMode.LstmOnly;
    Ocr.Language = OcrLanguage.EnglishFast;
    using (var Input = new OcrInput("..\\Images\\Purgatory.PNG"))
    {
        var Result = Ocr.Read(Input);
        Console.WriteLine(Result.Text);
    }
Dim Ocr = New IronTesseract()
	Ocr.Configuration.BlackListCharacters = "~`$#^*_}{][|\"
	Ocr.Configuration.PageSegmentationMode = TesseractPageSegmentationMode.Auto
	Ocr.Configuration.TesseractVersion = TesseractVersion.Tesseract5
	Ocr.Configuration.EngineMode = TesseractEngineMode.LstmOnly
	Ocr.Language = OcrLanguage.EnglishFast
	Using Input = New OcrInput("..\Images\Purgatory.PNG")
		Dim Result = Ocr.Read(Input)
		Console.WriteLine(Result.Text)
	End Using
VB   C#

Pricing and licensing options

There are essentially three paid licensing tiers that all work on a one-time purchase, lifetime license principle.

And yes, these are free for development purposes.

Further information

IronOCR features for .NET applications running OCR on Azure and other Systems

  • IronOCR supports 127 international languages. Each language is available in Fast, Standard and Best quality. Some of the language packs available include:
    • Bulgarian
    • Armenian
    • Croatian
    • Afrikaans
    • Danish
    • Czech
    • Filipino
    • Finnish
    • French
    • German
    • There are many more language packs available, to have a look at them, please follow the next link. IronOCR language packs
  • It works out of the box in .NET
    • Support for Xamarin
    • Support for Mono
    • Support for Microsoft Azure
    • Support for Docker on Microsoft Azure
    • Supports PDF documents
    • Supports Multiframe Tiffs
    • Support for all major image formats
  • The following .NET Frameworks are supported:
    • .NET Framework 4.5 and higher
    • .NET Standard 2
    • .NET Core 2
    • .NET Core 3
    • .NET Core 5
  • You don’t have to have Tesseract (an open-source ocr engine which supports Unicode and supports more than 100 languages) installed for IronOCR to work.
    • Has improved accuracy over Tesseract
    • Has improved speed over Tesseract
  • Corrects low quality scans of documents or files
  • Corrects low quality skewed scans of documents or files

What is (Optical Character Recognition) OCR?

According to Wikipedia: Optical character recognition is the electronic or mechanical conversion of images of typed, printed text into machine-encoded text, whether from a scanned document, a photo of a document, a scene-photo or from subtitle text superimposed on an image. Ocr stands for Optical Character Recognition. There are essentially four types of optical character recognition, they are:

  • OCR - Optical Character Recognition, that targets typewritten text from an input document, one character, or glyph (elemental symbol within an agreed set of symbols, for example ‘a’ in different fonts) at a time.
  • OWR - Optical Word Recognition, that targets typewritten text from an input document, one word at a time
  • ICR - Intelligent Character Recognition, that targets printed text such as print script (characters with no joining to other letters) and cursive text, one character or glyph at a time
  • IWR - Intelligent Word Recognition, that targets cursive text.