OCR Profile

Can be one of the following types:

The OCR Engine determines the core technology and algorithms used to extract text from images.

Usage

• Select the OCR Engine that best matches your document types and quality requirements.
• Different engines may offer varying levels of accuracy, speed, and support for languages or fonts.

Best Practices

• Test multiple OCR Engines on representative samples to determine which provides the best results for your use case.
• Some engines may have additional configuration options available as child properties or nodes.

The IP Profile defines a sequence of image processing operations (such as noise removal, line detection, or contrast adjustment) that are applied to the image before OCR is performed.

• These operations are applied only for the purpose of OCR and do not permanently alter the original image.
• Avoid including commands in the IP Profile that change the image size or resolution, as this can negatively impact OCR accuracy and downstream processing.

Typical Workflow

1. The IP Profile is applied to the image, producing a cleaned version for OCR.
2. The OCR Engine processes the cleaned image to extract text.
3. The original image remains unchanged in the repository.

Best Practices

• Use an IP Profile to improve OCR results on low-quality or noisy images.
• Review the steps in the IP Profile to ensure they are appropriate for your documents and do not introduce unwanted transformations.

Can be one of the following types:

When 'Image Segmentation' is enabled, the image is analyzed for regions that are bound on all four sides by lines or other separators.

• Each detected region is processed through OCR independently, which can improve accuracy on forms, tables, or documents with boxed layouts.
• This is especially useful for documents where information is presented in discrete regions, such as checkboxes, form fields, or grid cells.

Usage

• Enable 'Image Segmentation' for documents with complex line structures or when text is frequently enclosed in boxes.
• Adjust segmentation options as needed to fine-tune region detection and processing.

Best Practices

• Use in combination with synthesis and filtering options for optimal results on structured documents.
• Review the detected regions using diagnostic tools to ensure correct segmentation.

The 'Region of Interest' (ROI) restricts OCR processing to a specific rectangular area of the page, defined in inches.

• By default, the entire page is processed. Setting an ROI allows you to focus OCR on a particular section, such as a header, footer, or form field.
• The ROI is applied after any IP Profile cleanup and before OCR is performed.

Usage

• Use ROI to improve performance and accuracy when only a portion of the page contains relevant text.
• The ROI can be set visually using the zone editor in the UI, or by specifying coordinates directly.

Best Practices

• Ensure the ROI fully encompasses the desired content, accounting for possible shifts or variations in scanned images.
• Combine ROI with other filtering and segmentation options for precise extraction.

Can be one of the following values:

The 'Synthesis Method' determines whether Grooper will reprocess the output of the OCR Engine to improve the logical structure and flow of recognized text.

• When enabled, Grooper analyzes the spatial arrangement of recognized characters and lines, reconstructing text into more natural reading order and grouping.
• Synthesis can resolve issues such as fragmented lines, misaligned columns, or irregular spacing that are common in OCR output, especially from complex or tabular documents.

Usage

• Enable synthesis for documents where accurate text flow and structure are important, such as forms, tables, or multi-column layouts.
• Disabling synthesis may be appropriate when the OCR Engine already produces well-structured output, or when exact preservation of the engine's results is required.

Best Practices

• Test both settings on your document samples to determine which produces the most accurate and usable results.
• Synthesis is often recommended for documents with complex layouts or inconsistent text alignment.

Can be one of the following values:

'Font Pitch Mode' affects how Grooper inserts spaces and segments words during OCR synthesis, based on whether the text uses a fixed-pitch (monospaced) or variable-pitch (proportional) font.

• Auto: Grooper analyzes each segment to detect the most appropriate pitch mode, adapting to mixed or unknown fonts.
• Fixed: All characters are treated as having the same width, ideal for typewritten or monospaced documents.
• Variable: Character widths are allowed to vary, matching most modern printed or digital fonts.

Choosing the correct 'Font Pitch Mode' is important for accurate word segmentation, especially in documents with columns, tables, or inconsistent spacing.

Usage

• Use 'Auto' for most documents, especially when font type is unknown or varies within the document.
• Use 'Fixed' for legacy, typewritten, or monospaced text to improve segmentation accuracy.
• Use 'Variable' for documents known to use proportional fonts.

Best Practices

• Test different modes on your sample documents to determine which produces the best results.
• 'Auto' is generally recommended for flexibility and accuracy across diverse content.

'Maximum Variance' is used when 'Font Pitch' is set to 'Auto' and controls how strictly Grooper interprets a font as fixed-pitch (monospaced).

• A lower value requires character widths to be nearly identical to qualify as fixed-pitch.
• A higher value allows more variation before switching to variable-pitch processing.

Usage

• Decrease this value for documents with very consistent, typewritten text to enforce strict fixed-pitch detection.
• Increase for lower-quality scans or documents with minor distortions.

Best Practices

• The default value (0.12) works well for most scenarios.
• Adjust only if you observe incorrect word segmentation or space insertion due to pitch misclassification.

'Segment End Ratio' determines the threshold for splitting recognized text into separate segments or lines based on detected gaps.

• A higher value requires a larger gap to end a segment, resulting in longer lines or fewer splits.
• A lower value causes more frequent segmentation, which may be useful for tightly spaced or irregular text.

Usage

• Increase for documents with wide spacing between columns or fields to prevent premature line breaks.
• Decrease for documents with closely packed text or when over-segmentation is observed.

Best Practices

• The default value (1.0) is suitable for most documents.
• Adjust only if you notice lines being split incorrectly or merged inappropriately.

'Segment Reprocessing Threshold' enables automatic reprocessing of low-confidence text segments to improve recognition accuracy.

• Segments with average character confidence below this value are reprocessed through OCR.
• A value of 0 disables segment reprocessing entirely.

Usage

• Increase to aggressively reprocess more segments, which may improve results on difficult documents but increase processing time.
• Decrease to limit reprocessing to only the lowest-confidence segments.

Best Practices

• Use a moderate value (e.g., 0.5) for noisy or low-quality images.
• For clean, high-confidence documents, set to 0 to maximize performance.

'Iterations' enables iterative OCR, where the image is processed in multiple passes to improve recognition accuracy.

• On the first pass, OCR is performed normally.
• On subsequent passes, characters recognized in previous passes are digitally removed (dropped out), allowing the engine to focus on remaining or overlapping text.

Usage

• Set to 2 for documents with overlapping text, stamps, or noisy backgrounds that may obscure characters in a single pass.
• Use the default value (1) for most documents to maximize performance.

Best Practices

• Increasing iterations can improve results on challenging images but will increase processing time.
• Review results to ensure that additional passes do not introduce unwanted artifacts or duplicate characters.

'Cell Validation' is used to improve OCR accuracy on documents with columns, irregular layouts, or graphical elements that interfere with text flow.

• When enabled, the image is divided into a matrix of cells (see 'Rows' and 'Columns').
• OCR is performed on each cell, and results are merged with the full-page OCR.
• Edge characters within the buffer zone are automatically eliminated to reduce partial or overlapping text.

Usage

• Enable for forms, tables, or documents with non-standard text flow.
• Disable for simple, single-column documents.

Best Practices

• Adjust cell size and overlap to match the document's structure.
• Use diagnostic tools to visualize cell boundaries and validate results.

• Used only when 'Cell Validation' is enabled.
• Increasing the number of rows creates smaller vertical regions, which can help isolate text in multi-row layouts.

Configuration

• Set to match the number of logical rows in your document, or use a higher value for finer granularity.
• The default value (2) works for many forms and tables.

• Used only when 'Cell Validation' is enabled.
• Increasing the number of columns creates smaller horizontal regions, which can help isolate text in multi-column layouts.

Configuration

• Set to match the number of logical columns in your document, or use a higher value for finer granularity.
• The default value (2) is suitable for most columnar documents.

'Cell Edge Buffer' eliminates any characters overlapping the buffer zone at the edge of each cell.

• Helps remove partial or duplicated characters that may be detected on cell boundaries.
• Especially useful for documents with tightly packed or irregularly aligned text.

Usage

• Increase the buffer for noisy or low-quality images to reduce false positives.
• Decrease for documents where text is close to cell edges.

Best Practices

• The default value (0.1 inches) is effective for most scenarios.
• Ensure 'Cell Overlap' is greater than or equal to 'Cell Edge Buffer' to avoid gaps in coverage.

'Cell Overlap' ensures that text near the boundary between cells is not missed during OCR.

• Increasing overlap can improve accuracy for documents with text that crosses cell boundaries.
• Excessive overlap may increase processing time and the risk of duplicate characters.

Usage

• Set overlap to at least the value of 'Cell Edge Buffer' to ensure full coverage.
• Adjust based on the density and alignment of text in your documents.

Best Practices

• The default value (0.25 inches) is suitable for most forms and tables.
• Tune overlap and buffer together for optimal results.

'Skip First Column' can speed up cell validation by omitting the first column of cells from processing.

• Most cell validation issues occur when graphical elements appear to the left of the text.
• Skipping the first column is often sufficient for documents where the leftmost region does not contain relevant text.

Usage

• Enable to reduce processing time when the first column is not needed.
• Disable if important text may appear in the leftmost column.

Best Practices

• Review document samples to determine if skipping the first column is appropriate.
• Use in combination with cell diagnostics to verify coverage.

The 'Minimum Character Confidence' property filters out characters with low recognition confidence.

• Characters with confidence below this value are discarded from the results.
• A value of 0 disables character-level confidence filtering.

Usage

• Increase to remove uncertain or potentially incorrect characters from noisy or low-quality images.
• Decrease or set to 0 to retain all recognized characters, regardless of confidence.

Best Practices

• Use a moderate value (e.g., 0.5) for challenging documents.
• For clean, high-confidence images, a lower value may maximize data retention.

The 'Minimum Segment Confidence' property filters out entire text segments (such as words or lines) with low average character confidence.

• Segments with average confidence below this value are discarded.
• A value of 0 disables segment-level confidence filtering.

Usage

• Increase to remove unreliable segments from difficult or degraded documents.
• Decrease or set to 0 to retain all segments, regardless of confidence.

Best Practices

• Use in combination with 'Minimum Character Confidence' for optimal filtering.
• Review filtered results to ensure important data is not lost.

The 'Eliminate Edge Characters' property helps prevent partial or spurious characters at the boundaries of the ROI from being included in the OCR output.

• When enabled, any character closer than the specified minimum distance to the edge is discarded.

Usage

• Enable for documents where edge artifacts or partial characters are common.
• Disable if all text is well within the ROI boundaries.

Best Practices

• Adjust 'Minimum Distance to Edge' to balance between removing noise and retaining valid edge text.

The 'Minimum Distance to Edge' property works with 'Eliminate Edge Characters' to filter out characters too close to the ROI boundary.

• Characters within this distance from any edge are removed.

Usage

• Increase for noisy scans or when partial characters are frequently detected at the edges.
• Decrease for documents where valid text appears near the ROI boundary.

Best Practices

• The default value (0.01 inches) is suitable for most cases.
• Test with sample documents to ensure important data is not excluded.

The 'Eliminate Isolated Symbols' property helps filter out noise, marks, or artifacts that are not meaningful text.

• Segments are evaluated based on the 'Maximum Symbol Ratio' property.

Usage

• Enable for forms, checks, or documents with stray marks or isolated punctuation.
• Disable if symbols are important in your data.

Best Practices

• Adjust 'Maximum Symbol Ratio' to fine-tune symbol filtering.
• Review filtered output to ensure no required symbols are lost.

The 'Maximum Symbol Ratio' property determines what percentage of a segment's characters can be symbols before the segment is eliminated.

• A value of 1 requires all characters to be symbols for elimination.
• Lower values make the filter more aggressive.

Usage

• Decrease to remove segments with even a small proportion of symbols.
• Increase to allow more symbols in retained segments.

Best Practices

• The default value (1.0) is conservative; adjust as needed for your document type.

The 'Minimum Character Size' property removes characters smaller than the specified width or height.

• A value of 0 for width or height disables the minimum for that dimension.

Usage

• Increase to filter out specks, marks, or noise detected as tiny characters.
• Decrease to retain all small text.

Best Practices

• Set based on the smallest valid text in your documents.

The 'Maximum Character Size' property removes characters larger than the specified width or height.

• A value of 0 for width or height disables the maximum for that dimension.

Usage

• Decrease to filter out large blobs or artifacts.
• Increase to allow all large text.

Best Practices

• Set based on the largest valid text in your documents.

The 'Minimum Font Size' property removes characters with a measured font size smaller than the specified value.

• A value of 0 disables the minimum.
• 1 point = 1/72 inch.

Usage

• Increase to filter out very small, hard-to-read text.
• Decrease to retain all small text.

Best Practices

• Set based on the smallest font size you wish to capture.

The 'Maximum Font Size' property removes characters with a measured font size larger than the specified value.

• A value of 0 disables the maximum.
• 1 point = 1/72 inch.

Usage

• Decrease to filter out unusually large text or artifacts.
• Increase to allow all large text.

Best Practices

• Set based on the largest font size you wish to capture.

Can be one of the following types:

The 'Repair Options' property allows you to configure automated corrections for typical OCR mistakes, such as character substitutions or formatting issues.

• Use to improve accuracy and consistency in the final output.

Usage

• Configure as needed for your document types and error patterns.

Best Practices

• Review the available repair options and test their effect on your sample documents.

The 'Disable Junk Filtering' property controls whether Grooper attempts to remove stray marks, false punctuation, or other non-text artifacts.

• When false, non-letter characters are evaluated and discarded if determined to be junk.
• When true, all characters are retained, regardless of type.

Usage

• Enable for digitally generated documents or clean images with no stray marks.
• Disable for scanned images or documents prone to noise.

Best Practices

• Use in combination with other filtering options for optimal results.