OCR Output Types

Limitations of OCR

OCR software is also limited in what it is able to recognize. Most OCR software are only designed to recognize machine printed text, as opposed to handwriting. For handwriting there is ICR software (“Intelligent Character Recognition”). Desktop OCR applications include some limited ICR capabilities and can get acceptable accuracy with handprint. Cloud OCR solutions tend to get the best results for handwriting.

Similarly, most OCR software are only able to convert traditional machine fonts, not cursive scripts or calligraphy. There are many fonts out there, and OCR engines depend on common, separated letter shapes to recognize the text, so fonts that are unusual or flow together will not be recognized.

For more information, check these FAQs:

• Handprint Recognition guide
• Does ReadIRIS, FineReader or OmniPage support Zone OCR?
• Using OCR to capture data from tables and reports
• Can OCR be trained for specific fonts?

OCR Solutions for Business

OCR can do a lot more than convert scanned documents to Word and PDF files. Businesses can use OCR to automate a wide variety of document workflows and data entry tasks.

Business OCR data capture solutions including OCR servers for high volume conversions, document scanning and archiving systems, forms processing software with handprint recognition to capture surveys and applications, invoice processing for accounts payable automation, and document management systems to create secure repositories for searching, security and regulatory compliance.

Robotic Process Automation is becoming one of the most popular applications of OCR by making it possible for IT and knowledge workers to integrate OCR data capture into business workflows without having to write code or interface with APIs.

Integration services are available from our expert staff, each of whom has at least 10 years experience with implementing OCR data capture solutions for businesses.

Types of OCR Software

OCR Software for full-text conversion comes in many different types, which vary in price range based on their features, speed, and accuracy. OCR software for data capture is covered in another section.

For instance, you can get OCR freeware such as SimpleOCR or Tesseract that will serve in a pinch, but it will not provide acceptable accuracy if the document images are not pristine, and have other limitations like language support and the number of pages that can be processed at once.

One step up from freeware is Desktop OCR software. These are the best option if you need to convert several documents to Word or PDF and can spend $50-$100 to ensure that you get quality results with minimal need for corrections and reformatting.

If you have need to convert hundreds or thousands of documents, you can invest in a Batch OCR designed for scanning and converting large volumes of documents, or Server OCR software that watches “hot” folders for incoming documents in a variety of formats and languages and convert them to Word, PDF, eBook and other formats automatically.

For more information check out:

• Why are the prices of OCR applications so different?

Improving OCR Accuracy

Although some OCR engines are better than others, no software can guarantee 100% accuracy. This is because there are other factors in play, including scan quality. Recognition software will not be able to do its work if the scanner is not properly digitizing the page.

It is recommended to scan at a resolution of 300dpi for best results. Black & White (Bitonal) is preferred over Greyscale or Color modes, and although most modern scanners are fairly well configured out of the box, you may want to adjust your Brightness and Contrast settings for your particular documents.

If you do not have a scanner that has the necessary speed, quality, or other features that you require to scan your documents, you can always find a large selection of document scanners at ScanStore! ScanStore even has a handy scanners guide to help you find the perfect scanner for your specific requirements and price range.

For more on improving OCR accuracy check out these articles:

• What are the best scanner settings for OCR?
• Can OCR be trained for specific fonts?

The main features that differentiate OCR software are:

• Character recognition accuracy
• Page layout reconstruction accuracy
• Support for languages
• User interface design
• Output file formats (Word, Excel, PDF, eBook, etc.)
• OCR speed and support for multi-core CPUs
• Batch processing modes
• Advanced PDF encryption or compression
• Special features for niche projects

Because of the infinite combinations of document types, OCR engines, project requirements and special features, it may be possible that one engine will perform better with your particular documents than another. Use our handy OCR feature comparison chart to determine which OCR program best meets your requirements. And you can always ask an expert for a recommendation anytime!

What is the best ocr software?

Our OCR experts have tested the latest versions of FineReader, Kofax OmniPage, and ReadIRIS, and we consider ABBYY FineReader PDF the best overall value for business users, while ReadIRIS is the best ocr software for under $100.

The key deciding factors were:

• User interface design
• Page layout reconstruction capabilities
• Extensive language support
• Engine stability when processing large files
• Availability and quality of technical support

Though other testing labs have ranked OmniPage‘s overall accuracy slightly higher, we find the difference is nearly negligible. All modern OCR software has very good accuracy, so we recommend going with the one that has particular special features like ReadIRIS Corporate‘s CardIRIS, FineReader's camera OCR and screenshot reader, or OmniPage Ultimate's form data collection, auto-redaction and barcode filing capabilities.

If you would like to try them out yourself, you can download trial versions of ReadIRIS and FineReader from our store. Kofax does not provide demos for its OCR products.

Businesses with many documents to process should use our SimpleIndex batch document scanning software with the FineReader ocr engine to scan and OCR large batches of documents. Barcode and OCR can also be used to sort and file documents into folders, databases, SharePoint, and other cloud storage providers.

1. Scan or Upload Documents

Fortunately, many documents these days already exist as PDF files or JPEG images and don't need to be scanned. These files can be loaded into your OCR program directly.

If you're starting with paper, then first you have to get the documents ready. That means pulling any staples and paperclips, taping down loose edges, post-its, small documents and anything else that might get stuck in the document feeder.

In batch scanning scenarios, you may need to insert barcode separator sheets to indicate the start of each new document and automate filing.

You take these very neatly stacked piles of paper and feed them into the scanner. The neater the stacks, the less you have to open the scanner up and pull out little bits of paper and staples, and generally makes for a more pleasant and swear-free work environment.

Most OCR applications and modern scanners have very good image quality that require minimal configuration for good results, especially for color images. Black and White document images can be significantly smaller in file size, but it can be challenging to get consistently high-quality scans on variable paper types. 300 dots per inch resolution is recommended.

Free and desktop OCR applications will typically use a “Save As” style dialog to scan and save files one at a time. This is OK for a few documents, but if you have hundreds or thousands you'll want something more streamlined for batch scanning.

While capturing documents with digital camera is extremely convenient, it is not recommended for OCR unless you are using an application specifically designed to capture documents on your phone. While OCR technology has improved to compensate for the common distortions that cameras produce, it will never be as accurate as a scanned image.

2. Preprocessing and Classification

The first thing the OCR application does is to perform fast analysis on the image to identify document layout elements like pictures, paragraphs, tables, or barcodes.

For simple workflows there may only be a single document type. More complex batch document scanning or data capture applications that handle multiple types of documents must first identify the type of document based on the simple layout analysis so that specific OCR regions, languages, and other processing parameters can be correctly applied.

AI-based document classification incorporates user feedback to train and improve the layout analysis and more accurately identify document types as the system is used.

3. Perform the OCR Document Analysis

This is the step that reads the text on the document and performs the Optical Character Recognition.

This is an automated step that can take some time depending on what type of document is being read. Documents with handwriting, lots of small print text, tables, or background noise can significantly impact the processing speed.

For simple text conversion this can take less than a second per page, while more complex data extraction can take several seconds or longer.

4.Verify the OCR Results

Human verification of the text is required whenever errors are not acceptable. When converting large document archives to text primarily used for search, manual verification may not be required. For data capture applications where mistakes can lead to costly transaction errors, verification is essential.

OCR tools provide streamlined user interfaces that highlight words with possible errors, display the image and recognized text together, and let you quickly make corrections.

Verification is also how AI-based systems train their models to improve classification, field extraction, and text recognition accuracy.

In the past, OCR applications relied on user training of live samples to improve recognition rates. Modern OCR leverages huge datasets with billions of samples, so interactive training of the recognition patterns is no longer effective. When training is employed by modern data capture systems, it is focused on document type classification and field location training rather than the recognition of fonts and characters.

5. Export the Results to Your Apps

Now that you have your original document image and your converted OCR text, the final step is to put that data to use in whatever application it was that made you go down road in the first place.

OCR applications can export the documents and data to a variety of formats, the most common being:

• Searchable PDF files with OCR text hidden behind the image
• Word processing documents like DOCX or Google Docs
• Excel spreadsheets for documents formatted as tables
• Ebook reader formats like ePub and Mobi
• Structured text data files like CSV, XML, JSON, etc.
• SQL database servers
• Upload to cloud services via connector apps or APIs