Posts and articles addressing OCR accuracy and how to improve OCR results with optimal scanner settings, recognition parameters, and OCR-friendly document design.

Why are the prices of OCR applications so different?

OCR software ranges in price from freeware all the way up to tens of thousands of dollars. What explains the difference between these applications? Here’s the breakdown:

  • OCR Freeware uses the SimpleOCR or Tesseract engines and provide limited scanning and output format capabilities. Recognition quality is generally poor except for the highest quality document images.
  • PDF OCR Converters provide good quality OCR engines like ABBYY, IRIS and OmniPage, but limit the output to searchable PDF files. These cost less than $100.
  • Standard OCR applications range from $100-$200 and provide full OCR capabilities including converting scans to Word, Excel, HTML and other editable formats.
  • Corporate OCR applications add advanced features like automated hotfolder processing, concurrent licensing and other features useful for business applications. Pricing for these is $200-$500.
  • OCR Servers provide scalable, enterprise OCR services for processing very high volumes of documents or providing OCR capabilities to users throughout the organization. Prices start around $1,500 and go up based on processing volume.
  • Enterprise Data Capture and Forms Processing applications are used to capture structured data from complex documents like healthcare claim forms and invoices that include things like tables, handwriting, checkboxes, and movable zones. These solutions can cost anywhere from around $1,000 to hundreds of thousands of dollars depending on the document volume and complexity of the project.

Creating forms optimized for handprint recognition

Handprint recognition applications can provide dramatically different results in terms of accuracy depending on whether the form is designed with intelligent character recognition (ICR) in mind.

Forms Processing applications like ABBYY FlexiCapture have a built-in form design tool with ICR-optimized field layout elements and rules that validate whether your form uses best practices for recognition. These forms can be automatically converted to recognition templates for scanning for data capture. This saves you dozens of hours of trial and error during the design process and even more in data entry once the filled in forms are collected.

Best practice recommendations for ICR and OCR forms include:

  • Plenty of space between form elements and labels, at least 0.5cm / 0.25in
  • Use drop out colors for form backgrounds when possible
  • Hand printed characters should be constrained with boxes or combs to force filler to write legible, separated, printed characters
  • Use check boxes instead of handprint when possible since these are nearly 100% accurate
  • Use numeric codes instead of alphanumeric text when possible to reduce the number of possible characters and increase accuracy
  • Use validation rules to check against possible values and flag data with incorrect values
  • Check box fields can be used to verify the presence of signatures

Can OCR be trained for specific fonts?

OCR training was once a critical part of the conversion process. After a document was read, the operator would review the results to correct mistaken characters and these corrections would be used to train the engine so the next time you read a similar document the results are improved.

Modern OCR applications no longer rely on user training for accuracy unless you have very non-standard fonts. These engines have had decades of development and billions of samples used to train their algorithms. In most cases, the introduction of user training will only diminish the results for any documents that are different than the ones being trained.

The training functions still exist for these edge cases, but they are no longer an integral part of the OCR process.

Training in modern OCR is more likely to refer to enterprise data capture applications that use AI-based learning algorithms to find the locations of data points on documents with various different formats, such as invoices.

What are the best scanner settings for OCR?

Most OCR applications are optimized for 300 dots per inch resolution images.

While color is supported and most often performs better than black & white images, OCR algorithms will generally convert the color to B&W automatically as part of the OCR process. With color input, the dynamic conversion usually produces the best result, but not always.

Especially when an image contains stray markings, stamps, notes, colored paper or other elements that can throw off the binarization process, OCR results can be improved by paying careful attention to image processing settings and using a pristine black & white image for OCR instead of a color scan.

In forms processing and handprint recognition applications, guide marks in the form can often be removed during the scanning process, improving the OCR results when the software doesn’t have to distinguish between the form background and the words being recognized.

Using drop-out forms, traditionally printed in red or green and then scanned with a corresponding red or green light, automatically removes the form background during scanning and leaves only the text to be recognized. This can dramatically improve recognition results, especially for handprinted data.

Older, black & white scanners would require you to change out the lamps in order to perform color drop-out. All but the least expensive modern color scanners have the ability to enable drop-out colors in the scanner driver.

Advanced forms processing applications can perform color drop-out on-the-fly with scanned color images. Though this is generally not quite as accurate as scanning with a drop-out lamp enabled, it has the advantage of retaining a full-color original copy of the image with the form element and labels visible.

Go to Top