Image Segmentation Options

Configures how images are segmented into distinct regions for independent OCR processing.

Remarks

The Image Segmentation Options class provides advanced controls for dividing an image into separate regions—such as boxes, fields, or columns—so that each can be processed independently by OCR. This is especially useful for structured documents like forms, tables, or documents with boxed layouts, where text is often enclosed or separated by lines.

How Image Segmentation Works

When enabled, segmentation analyzes the image to detect regions that are visually bounded (e.g., by lines or whitespace). Each detected region is then processed as a separate OCR operation, which can improve recognition accuracy and prevent text from different regions from being merged incorrectly.

• Regions are identified based on boundaries, size, and area, using the configured options.
• Segmentation is typically used in conjunction with an OCR Profile, and can be enabled via the 'Image Segmentation' property.
• Detected regions are processed independently, and their results are merged into the final OCR output.

Configuration Guidance

• Minimum Size / Area: Use 'Minimum Size' and 'Minimum Area' to filter out small or irrelevant regions.
• Maximum Height / Width Ratio: Limit the size of detected regions to avoid merging unrelated content.
• Maximum Merge Height: Controls when adjacent regions are merged horizontally, useful for multi-line fields or table rows.
• Adjust these options based on the typical structure of your documents for optimal results.

Usage Scenarios

• Forms and Tables: Segmenting boxed fields or table cells to ensure each is recognized separately.
• Documents with Columns: Preventing text from different columns or regions from being combined.
• Complex Layouts: Handling documents with a mix of text, graphics, and lines.

Best Practices

• Enable segmentation for documents where text is frequently enclosed or separated by lines.
• Use diagnostic tools to visualize detected regions and fine-tune segmentation settings.
• Combine segmentation with synthesis and filtering options in your OCR Profile for best results on structured documents.
• Test on representative samples to ensure regions are detected and processed as intended.

Properties

Name	Type		Description
Minimum Size	String	►	The minimum width or height of a bounded region. ConverterSimple Value Converter Default6pt Specifies the smallest allowable dimension (width or height) for a detected region. Use this to filter out small boxes, marks, or noise that should not be processed as separate regions. The value can be set using common units (e.g., "6pt", "2mm"). Best Practice: Set this value slightly larger than the largest noise element on your documents to avoid false positives.
Minimum Area	String	►	The minimum area of a bound region. ConverterSimple Value Converter Default12pt Sets the minimum total area (width × height) required for a region to be considered valid. Use this to exclude very small regions that may result from specks, marks, or minor artifacts. The value can be set using area units (e.g., "12pt"). Configuration Tip: Increase this value if you see unwanted small regions being detected.
Maximum Height	String	►	The maximum height of a bound region. ConverterSimple Value Converter Default1in Limits the vertical size of detected regions. If set to zero or left empty, no maximum height is enforced. Useful for preventing the merging of unrelated vertical content, such as stacked fields or table rows. Usage: Set this to the typical height of your largest field or cell to avoid spanning multiple rows or sections.
Maximum Width Ratio	Double	►	The maximum width of a bound region, expressed as a percentage of the page width. ConverterPercent Converter Default75% Restricts the width of detected regions to a fraction of the total page width. Expressed as a decimal (e.g., 0.75 for 75% of the page width). Prevents regions from spanning across multiple columns or unrelated content. Best Practice: Lower this value for multi-column documents or forms to ensure each column or field is processed separately.
Maximum Merge Height	String	►	The maximum height of a bounded region which will be merged with its horizontal neighbors. ConverterSimple Value Converter Default14pt Controls when adjacent regions are merged together horizontally. Only regions with a height less than or equal to this value will be considered for merging. Useful for combining multi-line fields or table rows into a single region for OCR. Configuration Guidance: Set this to the typical height of a single row or field to merge only those that belong together.

Used By

OCR Profile