Identify provides the wrong ResolutionUnit for pyramid Tiff Files

[mexthecat]

Added space after %U in ImageMagick identify format prameter.

We work with pyramid Tiffs in our project. Since several images are combined in one file, identify provides all resolutions and resolution units one after the other. However, there is no space between the first and second image and the resolution unit is therefore read incorrectly.

ret = run(['identify', '-format', r'%[resolution.x] %[resolution.y] %U', img.filename], check=False, stderr=PIPE, stdout=PIPE)
...
tokens = ret.stdout.decode('utf-8').split(' ', 3)

['118.11', '118.11', 'PixelsPerCentimeter118.11', '118.11 PixelsPerCentimeter118.11 118.11 PixelsPerCentimeter118.11 118.11 PixelsPerCentimeter118.11 118.11 PixelsPerCentimeter']

A simple space after the %U should solve the problem

ret = run(['identify', '-format', r'%[resolution.x] %[resolution.y] %U ', img.filename], check=False, stderr=PIPE, stdout=PIPE)
...
tokens = ret.stdout.decode('utf-8').split(' ', 3)

['118.11', '118.11', 'PixelsPerCentimeter', '118.11 118.11 PixelsPerCentimeter 118.11 118.11 PixelsPerCentimeter 118.11 118.11 PixelsPerCentimeter 118.11 118.11 PixelsPerCentimeter ']

[bertsky]

Thanks a lot for the fix!

Do you happen to have an example image handy that we can use in the repo as a test case?

[mexthecat]

Yes - have some images but can not upload them here -> tif not allowed.

[bertsky]

Could you try to zip them and attach that?

(Also, is the image unproblematic for publication, i.e. no licensing or other legal problems?)

[mexthecat]

from https://archive.org/details/albumofindianfer00baynrich/albumofindianfer00baynrich/
0004.zip

[bertsky]

Turns out we already had an multi-frame TIFF example in the test assets – see 44deb80.

@kba what do you think, should we add 0004.tif as well? (It has cm instead of inches and should yield photometricInterpretation="YCBCR" but instead gives "RGB" because Pillow does not catch it and ImageMagick never gets asked about it.)

[bertsky]

Actually, Pillow just does not say anything. We could ask ImageMagick about it, though.

(However, %[EXIF:PhotometricInterpretation] does not yield anything... Not sure how to get that info.)

So IMO this test should state YCbCr and fail until we fix OcrdExif.run_identify to include the results from the CLI, or we find a way to make Pillow give us the info.

BTW, using Pillow's TiffTags we can actually get the right value:

# val = img.tag_v2.named()['PhotometricInterpretation']
val = img.tag_v2.get(262)
interpretation = dict((v, k) for k, v in TiffTags.TAGS_V2.get(262).enum.items())
val = interpretation[val]