Thumbs.db inverted reversed blue/red channels

[Sembiance]

Greetings

Thanks so much for providing the invert option, it really works great. However I have come across a Thumbs.db file where the invert worked great, except the red in the image comes out blue.

Attached is the Thumbs.db.zip

Here is the original image:

Here is the thumbnail extracted with the --invert option:

And here it is without --invert:

As you can see, --invert really helps a lot, but the blue taillights should be red. Perhaps the channel conversion is buggy in some way? Or maybe this is just the way the Thumbs.db format works and there isn't a solution.

Once again, thanks for creating Vinetto.

[AtesComp]

Yes, it does look like red is blue and blue is red. I can only speculate that the idiots at MS were (again) attempting to obfuscate an otherwise open format...JPEG...when they developed the thumbs.db format. I'm glad they got with the program in their later refactoring.

I'll look at updating the code. It may be that the images were matrix inverted which also swapped the red and blue channels. That would also account for the upside down, inverted image issues. This would fix all these perceived individual issues in one simple fix. I've been validating the image format in the binary encoding, but it's a PITA. If only I could find some actual documentation on this old format...

Thanks for finding this issue.

[AtesComp]

Oops.

[AtesComp]

As per issue #2, you can do the following to test.

image = Image.open( strMyImageFile )
r, g, b, a = image.split()
imageRGB = Image.merge("RGB", (b, g, r))

Note the red and blue transpose.

[AtesComp]

Thanks for the samples. I think I've resolutely fixed the Type 1 issues now that I've taken a deeper dive. The "invert" option is no longer needed. Check out the new release.

[Sembiance]

Just checked out the latest version, it works great! All the thumbs look spot on perfect now, no odd colorization issues. Thanks for digging into that and solving it!