How to Read Handwriting Images with IronOCR

Handwriting Image Input

:path=/static-assets/ocr/content-code-examples/how-to/read-handwritten-image.cs

using IronOcr;

// Instantiate OCR engine
var ocr = new IronTesseract();

// Load handwriting image
var inputHandWriting = new OcrInput();
inputHandWriting.LoadImage("handwritten.png");

// Perform OCR on the handwriting image
OcrHandwritingResult result = ocr.ReadHandwriting(inputHandWriting);

// Output the recognized handwritten text
Console.WriteLine(result.Text);
// Output the confidence score of the OCR result
Console.WriteLine(result.Confidence);

IRON VB CONVERTER ERROR developers@ironsoftware.com

Output

The ReadHandWriting method achieved a good 90.6% confidence score, correctly identifying the majority of the text and the main ideas, such as the opening phrase, "My name is Erin Fish."

This is a strong result because the input is a challenging handwritten script rather than simple printed text. While the engine did struggle with spacing and connected letters, the core message of the paragraph was successfully extracted. This shows that IronOCR is highly capable of handling complex, non-standard text.

ReadHandwriting Async Example

IronOCR additionally supports an asynchronous version of the ReadHandwriting method, ReadHandwritingAsync. This is extremely useful when dealing with asynchronous code that requires fetching the input image before running the operation. Using the same input from before, here is a code example demonstrating how to use the async method to output the same result:

:path=/static-assets/ocr/content-code-examples/how-to/read-handwritten-image-async.cs

using IronOcr;
using System.Threading.Tasks;

// Instantiate OCR engine
var ocr = new IronTesseract();

// Load handwriting image
var inputHandWriting = new OcrInput();
inputHandWriting.LoadImage("handwritten.png");

// Perform OCR using the async method with 'await'.
// The compiler automatically infers this top-level code block as an 'async Task Main()' method.
OcrHandwritingResult result = await ocr.ReadHandwritingAsync(inputHandWriting);

// Output the recognized handwriting text
Console.WriteLine(result.Text);
// Output the confidence score of the OCR result
Console.WriteLine(result.Confidence);

IRON VB CONVERTER ERROR developers@ironsoftware.com

You can also provide an integer as a second optional parameter, timeoutMs, to wait for a specified number of milliseconds before the operation automatically cancels. By default, this parameter has a value of -1, signifying there is no time limit and the operation will run until completion.

Reading Handwriting Images Challenges

Although IronOCR can achieve a high degree of confidence in retaining the overall structure and text, OCR as a whole still struggles with handwriting, leading to localized errors. Here are some of the common challenges that necessitate verification of the extracted output:

Irregular Spacing: When it comes to spacing in print, the distance between letters is uniform. However, the spacing between strokes and the way letters join in handwriting can vary greatly. This non-uniformity causes the OCR engine to incorrectly segment characters, leading to the errors shown above, where the word ununiformed is split into individual characters (u n u n i f o c m e d) instead of a single word.

Stroke variation: Each person has a unique handwriting style, and even a single person writes the same letter slightly differently each time. The way letters connect, and the overall pattern itself, vary significantly. This prevents the development of a "one-size-fits-all" model, as the engine must cope with the high variability in stroke slant, pressure, and form, making pattern matching inherently less reliable than with standardized fonts.

Ambiguous Character Shapes: Handwriting often uses simplified or hurried strokes, leading to ambiguous character shapes. For example, a quickly written 'e' might look like a 'c', or a connected 'l' and 'i' might be misidentified.

As such, when using this method, please ensure you verify the output matches the intended input, paying special attention to closely spaced or poorly formed words.