Possible bug reading JP2 as grayscale when should be in color

[fmw42]

Operating system

MacOS

Operating system, version and so on

Monterey

Description

The attached image seems to be coded improperly or there is a bug. Can you identify which?

Exiftool says:

 0 ImageHeight                     : 10000
    4 ImageWidth                      : 10000
    8 NumberOfComponents              : 4

but it also says

    3 ColorSpace                      : Grayscale

NumberOfComponents I interpret as having 4 channels, RGBI or RGBA, but then it says colorspace is grayscale. So I am confused about this file.

Steps to Reproduce

See above

Images

http://www.fmwconcepts.com/misc_tests/dop10rgbi_32_280_5652_1_nw_2021.jp2.zip

Original from first file at https://www.opengeodata.nrw.de/produkte/geobasis/lusat/dop/dop_jp2_f10/, but there are a lot of files there that likely have the same issue.

Most tools that I have used (such as GIMP and Photoshop and GraphicConverter) seem to read the file open it as grayscale (with or without) an alpha channel. I believe that is wrong. It should be an RGB with IR channel or perhaps alpha channel, not sure which.

Are these tools reading it with old JP2 software? Or is this a miscoding on the creator of the files? Or is this a bug in OpenJpeg code for JP2?

[rouault]

The issue is that the value of the EnumCS of the colr box is 17=GreyScale. It should rather be 16=sRGB

    <JP2Box name="colr" box_offset="66" box_length="15" data_offset="74" data_length="7">
      <BinaryContent>01000000000011</BinaryContent>
      <DecodedContent>
        <Field name="METH" type="uint8" description="Enumerated Colourspace">1</Field>
        <Field name="PREC" type="uint8">0</Field>
        <Field name="APPROX" type="uint8">0</Field>
        <Field name="EnumCS" type="uint32" description="greyscale">17</Field>
      </DecodedContent>
    </JP2Box>

[rouault]

by the way recent versions of Kakadu also complain about that file: "Looks like the file contains an illegal channel definitions (cdef) box; trying to fix the problem so that content can still be rendered."

[fmw42]

Agreed. It likely needs to be coded as RGB not grayscale. Also the 4th channel is coded as alpha and I think it might be an IR channel. The files are likely miscoded or they were created under an earlier definition of the format, though I suspect the former is the issue.