Question

I am using the Nuance OmniPage SDK to perform OCR on images and PDFs, for one of my cases I have to parse the XML file the OCR generates when recognizing a file, but the numbers don’t really make sense. Here’s a sample output for the SDK:

<page ocr-vers="OmniPageCSDK16" app-vers="OfficeDrop">
    <description backColor="ffffff">
        <source
            file="C:\Users\workspace\jobs\src\test\resources\test-docs\0003.tiff"
            dpix="300" dpiy="300" sizex="2480" sizey="3509" />
        <theoreticalPage size="Custom" marginLeft="0"
            marginTop="0" marginRight="0" marginBottom="0" width="2480" height="3509" />
    </description>
    <zones>
        <textZone l="1814" t="1517" r="4694" b="1733" fillingMethod="omnifont"
            recognitionModule="omnifontPlus2w" chrFilter="all">
            <ln l="1814" t="1517" r="4690" b="1728" baseLine="1680"
                underlined="none" fontSize="1100">
                <wd l="1814" t="1517" r="2539" b="1728">Sample</wd>
                <space width="67" />
                <wd l="2606" t="1541" r="2995" b="1680">text</wd>
                <space width="72" />
                <wd l="3067" t="1570" r="3302" b="1680">on</wd>
                <space width="72" />
                <wd l="3374" t="1517" r="3734" b="1680">line</wd>
                <space width="91" />
                <wd l="3826" t="1522" r="3912" b="1680">1</wd>
                <space width="82" />
                <wd l="3994" t="1570" r="4459" b="1728">page</wd>
                <space width="91" />
                <wd l="4550" t="1522" r="4690" b="1680">1.</wd>
                <space />
            </ln>
        </textZone>
    </zones>
</page>

The page size definition is quite easy to figure out, ((2480 * 72) / 300) is going to give me the correct page width in points (and then the height) for the final PDF page, but the other values don’t make any sense.

What is this font-size of 1100? And these word coordinates? How can they have coordinates that are in positions that do not exist on the page?

Is there anyone out there that knows what this XML format is or that could point me do documentation? I have searched around all manuals that came with the SDK and I still can’t find anything related to what these values represent or which unit they are using (i’m sure it isn’t pixels or points, i guess).

Any help is appreciated. I’m using java bindings to access the OmniPage C SDK.

Answer 1

Omni Page word coordidates are normally outputted in terms of Twips you need to do the respective point or pixel conversion. for example if I want to calculate measurement for a 200 dpi image I need to do this way (value*200)/1440 pixels .To cross check just open the image in paintbrush and see the pointer position for each word.