Optical Character Recognition (OCR) is the process of taking image-based versions of characters and converting them into machine encoded text. This enables your application to recognize and convert characters from scanned images or image-based documents into machine-readable, searchable text.
This capability is essential for transforming static visual content—such as scanned books, photos of documents, or handwritten forms—into usable digital text that can be indexed, selected, copied, or searched.
Some popular use cases include:
You have three OCR options to choose from when using OCR with the Apryse Server SDK:
The Apryse Server SDK offers a downloadable default OCR Module as an optional add-on utility in order to use OCR with the SDK. It is currently available on Windows, Linux, and macOS.
The default OCR Module is the newest OCR and delivers strong recognition capabilities across a wide range of document types. Beginning with version 12.0, it is based on modern Deep Learning Neural Networks for better accuracy and a wider variety of OCR language selection.
The default OCR module:
The previous V11 OCR Module is still available under the Alternative OCR Module section. For those who prioritize faster processing time on leaner hardware and less memory consumption, the alternative OCR Module may be the better choice.
For advanced layout scenarios—such as pages with multiple disconnected text regions like magazine covers or CAD drawings—you can optionally use the IRIS OCR Module, which may provide improved accuracy and layout interpretation. The IRIS module is available as an additional add-on for Windows and Linux platforms.
Using an OCR module, the SDK can create searchable and selectable text from images or PDFs, producing either a PDF with selectable text, or outputting just the text position data in reusable JSON or XML form.
The default OCR Module offers about 16% better word recognition accuracy due to its modern Deep Learning Neural Network-based implementation. It does not require a GPU to run.
The alternative and IRIS OCR modules offer 2-3x shorter processing time and less memory usage, ideal for those on limited hardware, or when speed is an absolute priority.
Once integrated, the OCR Module enables the SDK to generate Searchable PDFs with selectable text layers
The module takes advantage of pdftron.PDF.Convert.ToPdf internally and accepts multiple image formats, as well as PDFs with only raster images. The result quality depends on image supplied. The ideal image is color or grayscale with resolution in the vicinity of 300 DPI.
The default OCR includes models optimized for English-language scenarios, as well as multilingual models supporting over 80 languages.
Text orientation detection is not supported.
Text orientation detection is not supported for Korean language.
Korean language can’t be mixed with the rest of the group.
Text orientation detection is not supported.
Text orientation detection is not supported.
These languages can’t be mixed.
Note: You can include more than one language in the same document (for example, English + Arabic). The OCR engine can recognize mixed-language content as long as the selected languages are supported and follow the rules below.
Rules:
OCR workflow
In this section, we showcase the potential OCR workflow.
Did you find this helpful?
Trial setup questions?
Ask experts on DiscordNeed other help?
Contact SupportPricing or product questions?
Contact Sales