What's new

CSDK for Linux 2025.3 and its documentation now reflect Tungsten Automation Corporation as the company providing CSDK. Prior to January 2024, the company was known as Kofax, Inc.

As part of rebranding, the .NET namespace has been renamed from Kofax.OmniPageCSDK to OmniPageCSDK.

CSDK for Linux 2025.3 introduces the following Optical Character Recognition (OCR) enhancements:

• Improved OCR accuracy for Hebrew text.
• A new engine for Thai OCR has been introduced to improve accuracy and supportability. The module can recognize English words embedded in Thai text. There is no need to alter existing CSDK applications to leverage the new engine. You can revert to the previous module (KADMOS) by using the RM_RER recognition module setting. 
• Orientation detection for Hebrew and Thai text: CSDK can now automatically detect the orientation of Hebrew and Thai pages and rotate them as expected, resulting in improved OCR accuracy for documents in these languages.

• Improved OCR accuracy for Arabic text and Eastern Arabic numerals.

• OCR now recognizes strike-through fonts in Latin languages. This feature is supported for Office and HTML output formats while it is not supported for PDF output, as strike-through fonts are not defined in the PDF standard.

• Filter support for Asian characters: You can now specify character zones for Asian, Western, or numeric characters to improve OCR accuracy.

  For details, refer to the CHR_FILTER enum section in the RecAPI Online Help.

CSDK 2025.3 now includes improved OCR accuracy in machine-readable zones (MRZ) of passports, visas and ID documents.

FM_MRZ (Detection: MRZ) has been added as a new filling method in the FILLINGMETHOD enumeration. It allows zones to be configured for MRZ detection.

The corresponding LZX_MRZ flag has been introduced for extended page parsing. It can be used in combination with other LZX_* flags and passed to the kRecLocateZonesEx function to support the automatic locating of MRZs.

CSDK 2025.3 introduces advanced image pre-processing capabilities to enhance Optical Character Recognition (OCR) accuracy on scanned images, particularly ID cards and documents with challenging backgrounds, for example colored, watermarked, or patterned.

Key improvements include:

• The new Kernel.Img.Despeckle.ColorHalftone setting has been added to enable this additional processing. The default value of the setting is False.

• Noise removal and binarization: Reduces unwanted artifacts, for example spots while converting images to black-and-white, enabling clearer text recognition even on complex backgrounds.

• Adaptive binarization enhancements: Improves text detection on gray or patterned backgrounds where traditional binarization struggles. This enhancement minimizes issues and ensures text areas are preserved effectively, improving OCR accuracy on visually complex images.

CSDK for Linux 2025.3 now provides improved QR code detection and recognition.

CSDK for Linux 2025.3 now supports Code 32 barcodes, also called IMH, Italian Pharmacode and Radix 32 Barcode, the standard commissioned by the Italian Health Ministry.

CSDK for Linux 2025.3 now supports the hOCR and the ALTO open standards in the Direct TXT module to output OCR results.

For details, refer to the Tungsten OmniPage Capture SDK Online Help.

With CSDK for Linux 2025.3, the supported Python versions have been updated.

• Previous requirement: Python 3.6.xx, 3.7.xx, and 3.8.xx

• New requirement: Python 3.9, 3.10, 3.11, and 3.12

CSDK for Linux 2025.3 now provides a simple OmniPage Form Processor (OPFP) API to conduct data extraction.

OPFP is an add-on module to facilitate data extraction from problematic forms. It is efficient when template-based processing fails, as it can handle variable form layouts and data fields with no specific location. You can prepare a Form Descriptor containing field objects that define compound rules to identify the information on the form.

For details, refer to the Tungsten OmniPage Capture SDK Online Help.

CSDK for Linux 2025.3 introduces zone border settings for ICR (Intelligent Character Recognition) that can be defined to improve accuracy when recognizing machine printed zones and printed text handwritten in fillable forms.

For details, refer to the ZONE_BORDER_TYPE and kRecSetZoneBorder sections in the RecAPI Online Help.

CSDK for Linux 2025.3 now recognizes redacted areas in textual PDFs as text rather than anchored graphics and ensures redacted areas remain properly aligned with the surrounding content.

Previously, redactions were treated as static graphic elements, making it difficult to edit or add text near redacted content.

With this enhancement, redactions now behave as text and shift dynamically when text is inserted, edited, or removed before the redacted area. This improves text editing workflow in word processing applications like Microsoft Word, where redaction flexibility is essential.

CSDK for Linux 2025.3 has been updated to improve the detection and handling of Magnetic Ink Character Recognition (MICR) fonts.

Previously, MICR detection was limited to predefined zones, which posed challenges when using checks as document separators without defined MICR zones.

With CSDK for Linux 2025.3, you can now enable the MICR specific characters across the entire page, without falsely outputting MICR characters.

With CSDK for Linux 2025.3, the supported operating system versions have been updated.

• Previous requirement: Red Hat Enterprise Linux Server 7.x and 8

• New requirement: Red Hat Enterprise Linux Server 8.8, 9.x, and 10.x