DataFlex Requires Chilkat v11.7.0+

Extract the OCR Text Layer of a Scanned PDF

See more PDF Text Extraction Examples

Demonstrates a scanned page that carries an invisible OCR text layer over the page image. The quality report identifies the page's textSource as "ocrLayer". The "invisibleText" filter is "auto" by default, which uses the invisible text only on pages that have no visible text; "exclude" ignores it. DataFlex

Use ChilkatAx-win32.pkg

Procedure Test
    Boolean iSuccess
    Handle hoPdf
    Variant vJson
    Handle hoJson
    Variant vJsonOut
    Handle hoJsonOut
    String sTemp1
    String sTemp2

    Move False To iSuccess

    //  This example requires the Chilkat API to have been previously unlocked.
    //  See Global Unlock Sample for sample code.

    //  The PDF used in this example (a page image with an invisible text layer, as an OCR program produces) is available at scanned_with_ocr.pdf

    Get Create (RefClass(cComChilkatPdf)) To hoPdf
    If (Not(IsComObjectCreated(hoPdf))) Begin
        Send CreateComObject of hoPdf
    End
    Get ComLoadFile Of hoPdf "qa_data/pdf/scanned_with_ocr.pdf" To iSuccess
    If (iSuccess = False) Begin
        Get ComLastErrorText Of hoPdf To sTemp1
        Showln sTemp1
        Procedure_Return
    End

    Get Create (RefClass(cComChilkatJsonObject)) To hoJson
    If (Not(IsComObjectCreated(hoJson))) Begin
        Send CreateComObject of hoJson
    End
    Get Create (RefClass(cComChilkatJsonObject)) To hoJsonOut
    If (Not(IsComObjectCreated(hoJsonOut))) Begin
        Send CreateComObject of hoJsonOut
    End

    //  The page is a picture of a handwritten log with the OCR'd words drawn invisibly over it.
    //  "auto" (the default) includes invisible text exactly in this case: on a page with no visible text.
    Get pvComObject of hoJson to vJson
    Get pvComObject of hoJsonOut to vJsonOut
    Get ComExtractText Of hoPdf "text" vJson vJsonOut To iSuccess
    If (iSuccess = False) Begin
        Get ComLastErrorText Of hoPdf To sTemp1
        Showln sTemp1
        Procedure_Return
    End

    Get ComStringOf Of hoJsonOut "quality.pages[0].textSource" To sTemp1
    Showln "textSource: " sTemp1
    Get ComStringOf Of hoJsonOut "quality.pages[0].rating" To sTemp1
    Showln "rating: " sTemp1
    Get ComStringOf Of hoJsonOut "text" To sTemp1
    Showln sTemp1

    //  The quality report flags the page, because OCR text is only as good as the OCR was.
    Get ComStringOf Of hoJsonOut "quality.pages[0].issues[0].code" To sTemp1
    Get ComStringOf Of hoJsonOut "quality.pages[0].issues[0].message" To sTemp2
    Showln "issue: " sTemp1 " - " sTemp2

    //  "exclude" ignores invisible text altogether: the page then has no text at all.
    Get ComUpdateString Of hoJson "invisibleText" "exclude" To iSuccess
    Get pvComObject of hoJson to vJson
    Get pvComObject of hoJsonOut to vJsonOut
    Get ComExtractText Of hoPdf "text" vJson vJsonOut To iSuccess
    If (iSuccess = False) Begin
        Get ComLastErrorText Of hoPdf To sTemp1
        Showln sTemp1
        Procedure_Return
    End

    Get ComStringOf Of hoJsonOut "quality.pages[0].textSource" To sTemp1
    Get ComStringOf Of hoJsonOut "text" To sTemp2
    Showln "With invisibleText excluded, textSource: " sTemp1 ", text: [" sTemp2 "]"

    //  Sample output:

    //  textSource: ocrLayer
    //  rating: good
    //  Keeper's Log, 12 October 1902
    //  
    //  Wind north-west, fresh. Lamp lit at 5.40 p.m. and burning well all night.
    //  
    //  Schooner Mary Ann passed inbound at 9 p.m. showing proper lights.
    //  
    //  Oil on hand: forty-one gallons. Cleaned the lens at dawn.
    //  issue: ocrLayer - The text is an invisible OCR layer over an image
    //  With invisibleText excluded, textSource: ocrLayer, text: []


End_Procedure