Canon ColorData4: for ColorDataVersion=3, SpecularWhiteLevel is most likely at 617

[LebedevRI]

  if(f == ColorDataFormat::ColorData4 && ver == 3) {
    for(unsigned i = 0x0e7; i != wb->count; ++i) {
      auto val = wb->getU16(i);
      if(val >= 16384>>2 && val <= 16384)
        writeLog(DEBUG_PRIO::ERROR, "i=%04i: %i", i, val);
    }
    ThrowRDE("ver=3");
  }

~/raw-camera-samples/raw.pixls.us/Canon$ ~/rawspeed/build-Clang17-release/src/utilities/rstest/rstest EOS\ 40D/*.CR2
EOS 40D/_MG_0153.CR2                                   : starting decoding ... 
EOS 40D/_MG_0154.CR2                                   : starting decoding ... 
RawSpeed:i=0586: 7899
RawSpeed:i=0602: 9374
RawSpeed:i=0617: 13481
RawSpeed:i=0629: 5826
RawSpeed:i=0634: 4195
RawSpeed:i=0637: 4262
RawSpeed:i=0657: 8191
RawSpeed:i=0659: 8191
EOS 40D/_MG_0153.CR2                                   :  10 MB /   65 ms
RawSpeed:i=0586: 8137
RawSpeed:i=0602: 9366
RawSpeed:i=0617: 13450
RawSpeed:i=0629: 5826
RawSpeed:i=0634: 4195
RawSpeed:i=0637: 4262
RawSpeed:i=0657: 8191
RawSpeed:i=0659: 8191
EOS 40D/_MG_0154.CR2                                   :   5 MB /   80 ms
EOS 40D/_MG_0153.CR2 failed: hash/metadata mismatch
EOS 40D/_MG_0154.CR2 failed: hash/metadata mismatch
Total decoding time: 0.145s

WARNING: the following 2 tests have failed:
EOS 40D/_MG_0153.CR2 failed: hash/metadata mismatch
EOS 40D/_MG_0154.CR2 failed: hash/metadata mismatch
See rstest.log for details.

13481 and 13450 seem within ballpark for the specular white level for those files.

[LebedevRI]

@boardhead

diff --git a/lib/Image/ExifTool/Canon.pm b/lib/Image/ExifTool/Canon.pm
index f3ec2434..e8a775ba 100644
--- a/lib/Image/ExifTool/Canon.pm
+++ b/lib/Image/ExifTool/Canon.pm
@@ -7494,16 +7494,21 @@ my %ciMaxFocal = (
     0x280 => { #PH
         Name => 'RawMeasuredRGGB',
         Format => 'int32u[4]',
         Notes => 'raw MeasuredRGGB values, before normalization',
         # swap words because the word ordering is big-endian, opposite to the byte ordering
         ValueConv => \&SwapWords,
         ValueConvInv => \&SwapWords,
     },
+    0x269 => {
+        Name => 'SpecularWhiteLevel',
+        Condition => '$$self{ColorDataVersion} == 3',
+        Format => 'int16u',
+    },
     0x2b4 => { #IB
         Name => 'PerChannelBlackLevel',
         Condition => '$$self{ColorDataVersion} == 4 or $$self{ColorDataVersion} == 5',
         Format => 'int16u[4]',
     },
     0x2b8 => { #IB
         Name => 'NormalWhiteLevel',
         Condition => '$$self{ColorDataVersion} == 4 or $$self{ColorDataVersion} == 5',

[LebedevRI]

And it looks like that blackLevel[4] is only at 231 (0x0E7), there are no other black levels for that format+version.

[LebedevRI]

(@boardhead note that this is not a duplicate of #232.)

[boardhead]

Thanks for this. However, this seems to give inconsistent results for the 40D samples I have on hand:

exiftool ../pics/Canon*40D* -specularwhitelevel -if 'defined $specularwhitelevel' -H -firmwareversion
======== ../pics/Canon40D-1.0.4.jpg
0x0269 Specular White Level            : 16113
0x00ff Firmware Version                : 1.0.4
======== ../pics/CanonEOS40D.CR2
0x0269 Specular White Level            : 19
0x00ff Firmware Version                : 1.0.3
======== ../pics/CanonEOS40D.jpg
0x0269 Specular White Level            : 145
0x00ff Firmware Version                : 1.0.3
======== ../pics/CanonEOS40D_firmware1.0.4.CR2
0x0269 Specular White Level            : 45
0x00ff Firmware Version                : 1.0.4
======== ../pics/CanonEOS40D_firmware1.0.5.CR2
0x0269 Specular White Level            : 228
0x00ff Firmware Version                : 1.0.5
======== ../pics/CanonEOS40D_sRAW.CR2
0x0269 Specular White Level            : 816
0x00ff Firmware Version                : 1.0.3
======== ../pics/CanonEOS40D_sRAW.dng
0x0269 Specular White Level            : 816
0x00ff Firmware Version                : 1.0.3
    7 image files read

• Phil

[LebedevRI]

Oh, nice catch!
Looking at more samples, i do agree that it's not 0x0269.
What about 0x0275?

[boardhead]

For all my 40D samples 0x0275 is less than 7300, but for all my other Canon samples the value is greater than 9990, so I don't think this is it.

• Phil

[LebedevRI]

https://discuss.pixls.us/t/any-canon-40d-users-around-a-bit-of-testing-wanted/40843

[cytrinox]

Based on my research 40D Colordata has no whitepoint information. I've hardcoded whitepoint to 13600 (iirc source was dcraw.c).
https://github.com/dnglab/dnglab/blob/main/rawler/data/cameras/canon/40d.toml

Edit: FYI Reference to colordata implementation in dnglab: https://github.com/dnglab/dnglab/blob/main/rawler/src/decoders/cr2/colordata.rs#L35

[LebedevRI]

Ok, looking at some more samples (@MStraeten, thank you!),
i'm going to agree that it does not seem like there's SpecularWhiteLevel
in that camera's MakerNotes (and thus, likely in any earlier cameras too).

Thank you all!

Roman