Skip to main content

Version: Version 1.0.0

Engine parameters

The Pdftools OCR Service lets you fine-tune the OCR engine through various parameters influencing performance, layout detection, output quality, and preprocessing steps. You can set these engine parameters as a sequence of key and value pairs separated by a semicolon (;).

General settings

`PredefinedProfile`

Key	Type	Default
`PredefinedProfile`	Name	`Default`

The name of the predefined recognition profile. Predefined profiles offer optimized OCR settings for different goals. You can pass them using the PredefinedProfile parameter.

Profile name	Purpose
`DocumentConversion_Accuracy`	Maximizes accuracy in text recognition.
`DocumentConversion_Speed`	Optimizes for processing speed.
`DocumentArchiving_Accuracy`	Preserves document structure and layout.
`DocumentArchiving_Speed`	Fast archiving with basic layout.
`BookArchiving_Accuracy`	High-quality capture for books.
`BookArchiving_Speed`	Speeds up book digitization.
`TextExtraction_Accuracy`	For extracting clean text content.
`TextExtraction_Speed`	Fast plain-text extraction.
`FieldLevelRecognition`	Fine-tuned for form field recognition.
`BarcodeRecognition_Accuracy`	High accuracy for barcode content.
`BarcodeRecognition_Speed`	Fast barcode detection.
`EngineeringDrawingsProcessing`	Optimized for linework and schematics.
`Default`	General-purpose balanced settings.

`DocumentConversion_Accuracy`

Convert documents into editable formats, optimized for accuracy. Enables font style detection.

This profile corresponds to the following parameter:¹

[BarcodeParams]
EnableAdvancedExtractionMode = TRUE

`DocumentConversion_Speed`

For converting documents into editable formats, optimized for speed.

Similar to DocumentConversion_Accuracy, but document analysis and recognition are faster.

This profile corresponds to the following parameters:

[ObjectsExtractionParams]
ProhibitColorImage = TRUE

[PrepareImageMode]
UseFastBinarization = TRUE

[RecognizerParams]
FastMode = TRUE

`DocumentArchiving_Accuracy`

For creating an electronic archive, optimized for accuracy:

Enables detection of maximum text on an image, including text embedded into the image.
Skew correction is not performed.
Fonts and styles are not detected.

This profile corresponds to the following parameters:

[BarcodeParams]
EnableAdvancedExtractionMode = TRUE

[ObjectsExtractionParams]
DetectTextOnPictures = TRUE
EnableAggressiveTextExtraction = TRUE

[PageAnalysisParams]
EnableTextExtractionMode = TRUE

[SynthesisParamsForDocument]
DetectDocumentStructure = FALSE
DetectFontFormatting = FALSE

[SynthesisParamsForPage]
AllowGrayTextColor = TSPV_Yes
AllowGrayBackgroundColor = TSPV_Yes
DetectFontFormattingAtPageLevel = TRUE
DetectTextColor = TSPV_Yes

`DocumentArchiving_Speed`

Like DocumentArchiving_Accuracy, but document analysis and recognition are sped up.

This profile corresponds to the following parameters:

[ObjectsExtractionParams]
DetectMatrixPrinter = FALSE
DetectPorousText = FALSE
DetectTextOnPictures = TRUE
EnableAggressiveTextExtraction = TRUE
FastObjectsExtraction = TRUE
ProhibitColorImage = TRUE
RemoveGarbage = TRUE
RemoveTexture = FALSE

[PageAnalysisParams]
EnableTextExtractionMode = TRUE
ProhibitModelAnalysis = TRUE

[PrepareImageMode]
CorrectSkew = FALSE
UseFastBinarization = TRUE

[RecognizerParams]
FastMode = TRUE

[SynthesisParamsForDocument]
DetectDocumentStructure = FALSE
DetectFontFormatting = FALSE

[SynthesisParamsForPage]
AllowGrayBackgroundColor = TSPV_Yes
AllowGrayTextColor = TSPV_Yes
DetectFontFormattingAtPageLevel = TRUE
DetectTextColor = TSPV_Yes

`BookArchiving_Accuracy`

For creating an electronic library, optimized for accuracy:

High quality. Enables font style detection.

This profile corresponds to the following parameters:

[BarcodeParams]
EnableAdvancedExtractionMode = TRUE

`BookArchiving_Speed`

For creating an electronic library, optimized for speed:

Like BookArchiving_Accuracy, but document analysis and recognition are faster.

This profile corresponds to the following parameters:

[ObjectsExtractionParams]
ProhibitColorImage = TRUE

[PrepareImageMode]
UseFastBinarization = TRUE

[RecognizerParams]
FastMode = TRUE

`TextExtraction_Accuracy`

For extracting text from documents, optimized for accuracy:

Enables detection of all text on an image, including small text areas of low quality (pictures and tables are not detected).
Fonts and styles are not detected.

This profile corresponds to the following parameters:

[BarcodeParams]
EnableAdvancedExtractionMode = TRUE

[ObjectsExtractionParams]
DetectTextOnPictures = TRUE
EnableAggressiveTextExtraction = TRUE

[PageAnalysisParams]
DetectPictures = FALSE
EnableTextExtractionMode = TRUE

[SynthesisParamsForDocument]
DetectDocumentStructure = FALSE
DetectFontFormatting = FALSE

[SynthesisParamsForPage]
DetectFontFormattingAtPageLevel = TRUE

`TextExtraction_Speed`

For extracting text from documents, optimized for speed:

Like TextExtraction_Accuracy, but document analysis and recognition are sped up.

This profile corresponds to the following parameters:

[ObjectsExtractionParams]
DetectMatrixPrinter = FALSE
DetectPorousText = FALSE
DetectTextOnPictures = TRUE
EnableAggressiveTextExtraction = TRUE
FastObjectsExtraction = TRUE
ProhibitColorImage = TRUE
RemoveGarbage = TRUE
RemoveTexture = FALSE

[PageAnalysisParams]
DetectPictures = FALSE
EnableTextExtractionMode = TRUE
ProhibitModelAnalysis = TRUE

[PrepareImageMode]
CorrectSkew = TRUE
DiscardColorImage = TRUE
UseFastBinarization = TRUE

[RecognizerParams]
FastMode = TRUE

[SynthesisParamsForDocument]
DetectDocumentStructure = FALSE
DetectFontFormatting = FALSE

[SynthesisParamsForPage]
DetectFontFormattingAtPageLevel = TRUE
DetectTextColor = TSPV_Yes

`FieldLevelRecognition`

For recognizing short text fragments.

This profile corresponds to the following parameters:

[DocumentProcessingParams]
PerformSynthesis = FALSE

[PageProcessingParams]
PerformAnalysis = FALSE

[SynthesisParamsForPage]
DetectFontFormattingAtPageLevel = FALSE

`BarcodeRecognition_Accuracy`

For barcode extraction, optimized for accuracy:

Extracts only barcodes (text, pictures, or tables are not detected).

This profile corresponds to the following parameters:

[BarcodeParams]
MinRatioToTextHeight = 0.9

[ObjectsExtractionParams]
DetectMatrixPrinter = FALSE
DetectPorousText = FALSE

[PageAnalysisParams]
DetectBarcodes = TRUE
DetectPictures = FALSE
DetectTables = FALSE
DetectText = FALSE
DetectSeparators = FALSE
DetectVectorGraphics = FALSE

[PrepareImageMode]
CorrectSkew = FALSE

`BarcodeRecognition_Speed`

For barcode extraction, optimized for speed:

Like BarcodeRecognition_Accuracy, but document analysis and recognition are sped up.

This profile corresponds to the following parameters:

[BarcodeParams]
MinRatioToTextHeight = 0.9

[ObjectsExtractionParams]
DetectMatrixPrinter = FALSE
DetectPorousText = FALSE
FastObjectsExtraction = TRUE

[PageAnalysisParams]
DetectBarcodes = TRUE
DetectPictures = FALSE
DetectTables = FALSE
DetectText = FALSE
DetectSeparators = FALSE
DetectVectorGraphics = FALSE

[PageProcessingParams]
PerformPreprocessing = FALSE

[PrepareImageMode]
CorrectSkew = FALSE
DiscardColorImage = TRUE

`EngineeringDrawingsProcessing`

For recognizing technical drawings:

It takes into account large size and complexity of engineering diagrams, as well as the possibility of different text orientation within the image.
Enables detection of all text on an image, including text blocks of vertical orientation.

This profile corresponds to the following parameters:

[PageAnalysisParams]
DetectPictures = FALSE
DetectVectorGraphics = FALSE
DetectVerticalEuropeanText = TRUE
EnableTextExtractionMode = TRUE

[SynthesisParamsForDocument]
DetectDocumentStructure = FALSE
DetectFontFormatting = FALSE

[SynthesisParamsForPage]
DetectFontFormattingAtPageLevel = TRUE

`Default`

For default values:

Sets all the processing parameters to the default values.

`Profile`

Key	Type
`Profile`	Path

Path to a custom recognition profile .ini file.

Use the Profile parameter to reference a custom OCR profile that defines specific processing behavior. The file must be on the same machine as the Pdftools OCR Service Manager node. If both Profile and PredefinedProfile are set, the custom Profile overrides the predefined profile. The following snippet represents an example path to the custom profile .ini file in the Profile parameter:

Profile="C:\ocr\profiles\document_conversion_high_accuracy.ini"

Custom Profiles

For more details about creating your profiles, review Custom Profiles page.

`PreprocessingOnly`

Key	Type	Default
`PreprocessingOnly`	Boolean	`false`

If you enable the PreprocessingOnly, only image transformations (for example, deskewing, resolution correction, binarization) are applied, and no recognition is performed. This is useful for workflows that require cleaned-up images without OCR. The PreprocessingOnly parameter takes Boolean values true or false.

`RemoveGarbage`

Key	Type
`RemoveGarbage`	Integer

Remove small, isolated dark regions in bitonal images that are likely scanning noise before any OCR is done. The value defines the maximum area of such noise in pixels. A value of -1 enables automatic determination.

Blank page detection

`RecognizeBlankPages`

Key	Type	Default
`RecognizeBlankPages`	Boolean	`false`

Enable automatic skipping of pages that are considered blank. A blank page is a page with uniform coloring and only slight noise. Colored, grayscale, and bitonal pages can be subject to blank page recognition. If a page is skipped as blank, no OCR is performed. The RecognizeBlankPages parameter takes Boolean values true or false.

`BlankPageMargin`

Key	Type	Default
`BlankPageMargin`	Double	`0.02`

Set the ratio that the margin takes with respect to the corresponding page length. The margin is excluded from the analysis when a page is blank, preventing border artifacts from affecting blank page detection. Allowed values range from 0.0 to 0.5. This parameter is only active if the value of RecognizeBlankPages is set to true.

Output format control

`DisableMaskEmbedding`

Key	Type	Default
`DisableMaskEmbedding`	Boolean	`false`

If this option is set to true, no mask is embedded in the output TIFF. When set to false (default), bitonal masks are embedded in the output TIFF or PDF as an image layer. The mask layer is omitted when enabled, and only recognized text is preserved. The DisableMaskEmbedding is useful for output without background images. The DisableMaskEmbedding parameter takes Boolean values true or false.

All occurences of sentence “profile corresponds to the following parameters” reference a custom profile .ini file with an equivalent configuration to a predefined profile. For more information, review Profile, and Custom Profiles. ↩