OCR Cleanup

Performs temporary image cleanup to improve OCR results by removing non-text features.

Remarks

The OCR Cleanup command preprocesses images for OCR by removing content that typically interferes with text recognition, such as lines, barcodes, halftone patterns, specks, and checkboxes. This command is designed for use in OCR workflows where the goal is to maximize text recognition accuracy by temporarily removing non-text features.

The cleanup pipeline executes multiple stages:

1. Halftone Removal: Detects and removes halftone patterns that can confuse OCR engines.
2. Hole Punch Removal: Removes circular artifacts from hole punches.
3. Small Speck Removal: Eliminates small isolated dots and noise.
4. Barcode and Box Removal: Detects and removes barcodes and box-like features.
5. Large Speck Detection: Identifies and removes larger specks using both morphological and watershed techniques.
6. Text Segment Detection: Segments and preserves regions likely to contain text, protecting them from dropout.
7. Line Removal: Removes horizontal and vertical lines that may interfere with text recognition.
8. Checkbox Detection (optional): If enabled, detects and removes checkboxes, which are common in forms.

The command applies a combination of binarization, dropout, and feature detection to isolate text. It supports separate binarization strategies for low-contrast images, automatically switching to more aggressive or adaptive thresholding when needed.

Configuration and Usage

• Use 'Dropout Method' to control which non-text features are removed and how dropout regions are filled.
• Configure 'Binarization' and 'Low Contrast Binarization' to optimize text preservation and feature removal for different image types.
• Enable 'Detect Check Boxes' to remove checkboxes from forms and surveys.
• Always review the cleaned image to ensure important content is not lost.

Supported Pixel Formats

• Pixel8bppGrayscale
• Pixel24bppBgr
• Pixel1bppIndexed

Images are automatically converted as needed for processing.

Diagnostics

When run in diagnostic mode, OCR Cleanup generates diagnostic images and logs for each stage of the cleanup pipeline:

• Binarized: Shows the effect of thresholding and preprocessing.
• After Negated Region Removal: Displays the image after negative region cleanup.
• Dropout Mask: Visualizes the regions that will be removed.
• Log Messages: Reports feature detection, processing steps, and timing.

Use these diagnostics to fine-tune detection parameters and ensure that only the intended features are removed.

Notes

• This command is intended for temporary cleanup in OCR workflows; it should not be used for archival image cleanup.
• Use this command as an altenative to building a custom IP Profile for OCR preprocessing.
• The cleanup pipeline is designed to maximize OCR accuracy by removing only those features that are likely to interfere with text extraction.
• OCR Cleanup does not generate classification features directly, but the results can impact downstream OCR and extraction.

Properties

Name	Type		Description
General
Dropout Method	Dropout Method	►	Specifies the method used to remove non-text objects from the image. ConverterType Selector Requiredtrue Can be one of the following types: Value Description Fill Fills masked regions of an image with a solid color. Inpaint Digitally restores masked regions of an image by filling in missing or damaged areas. The 'Dropout Method' property determines how regions identified as non-text (such as lines, boxes, specks, or checkboxes) are removed or replaced in the image. Choose a dropout method to target specific feature types (e.g., lines, boxes, specks). The selected method determines how the dropout mask is generated and applied, and how the removed regions are filled (e.g., with white, background color, or interpolation). For most OCR workflows, "White Fill" is recommended to ensure a clean background for text recognition. Review the Dropout Mask diagnostic image to verify that only unwanted features are being removed.
Binarization	Binarize	►	The default binarization settings to use for most images. ConverterExpandable Converter The 'Binarization' property controls how the image is converted to black and white before feature removal. Adjust thresholding and adaptive options to best preserve text and remove noise. For most standard documents, "Auto" or "Adaptive" methods provide the best results. Use the Binarized diagnostic image to verify that text is preserved and non-text features are clearly separated.
Low Contrast Binarization	Binarize	►	The binarization settings to use for low-contrast images. ConverterExpandable Converter The 'Low Contrast Binarization' property is used automatically when the image is detected as low contrast. Allows more aggressive or adaptive thresholding to extract faint or degraded text. Configure this property to handle images where standard binarization fails to separate text from the background. Use the Binarized diagnostic image to verify that faint text is extracted and noise is minimized.
Detect Check Boxes	Boolean	►	If enabled, detects and removes checkboxes from the image. DefaultTrue The 'Detect Check Boxes' property controls whether checkboxes are detected and removed during cleanup. Enable this property to clean up forms and surveys prior to OCR, removing checkboxes that may interfere with text recognition. Only affects features identified as checkboxes; other content is preserved. Review the cleaned image to ensure that only checkboxes are removed and important content is retained.
Command Info
Command Type	String	►	The display name of this image processing command. The 'Command Type Name' property shows the user-friendly name of the IP Command as it appears in the Grooper UI. This helps users identify which command is being configured or executed, especially when working with IP Profiles containing multiple commands. This property is read-only and is automatically set based on the command's type.
Supported Pixel Formats	String	►	The pixel formats supported by this command. The 'Supported Pixel Formats' property lists the image pixel formats that the IP Command can process. If an image does not match one of these formats, the command will be skipped for that image. By default, all pixel formats are supported unless a derived command restricts this list. Example formats include Pixel8bppGrayscale, Pixel24bppBgr, and Pixel1bppIndexed. Use this property to verify compatibility when building IP Profiles or troubleshooting why a command did not execute.

See Also

FillInpaintBinarize

Used By

IP Step