Hi,

I've been experimenting with IPTC additions (under OS X) to Canon EOS 350D (aka: Digital Rebel XT) RAW files (.CR2 and it's all very simple: thanks !) and it looks like the Canon supplied RAW tweaking software Digital Photo Professional 1.6.1 (DPP) creates extra data that ExifTool can't understand and so removes when writing back to the edited RAW file.

For those without the software, DPP allows colour alterations using curves as well as other histogram based alterations, cropping, simple dust retouching and the like, but it does this non-destructively on the RAW file by applying what Canon term a 'Recipe' to the .CR2 file. DPP computes a changed preview thumbnail for viewing within the main app browser window but applies the steps in the recipe to the original file before allowing more edits. The changes can then be applied in batch elsewhere, but again all changes are non-destructive with respect to the original data.

If I add an IPTC tag before retouching, then everything is fine. If I retouch and then add a tag, DPP looses the changes and thinks the file is fresh from the camera. A quick glance at the file sizes shows that a far chunk of information (500KB or so) has been removed in this process, but DPP still understands the file and nothing crashes.

This makes me think that the extra information applied by DPP is done so in a sane fashion that is correctly tagged in terms of size etc., but what are the next steps I can take in order to find out the essential header data to allow ExitTool to leave the DPP block intact ?

I'm pretty happy at the command line and perl doesn't scare me (although my code may scare others !) but if anyone has some pointers for how to go about investigating this then I'd be most grateful. If others wish to look then I can provide any number of example images (but they are all aorund 8MB in size).

TTFN,
--
Ian Spray

I now understand what DPP is doing. It is simply appending a block of data to the end of the file without otherwise modifying the file at all. This makes a certain amount of sense, and explains why the data isn't referenced from within the TIFF structure. But this is a bit of a pain to say the least because TIFF is not a sequential format, so it is not as easy as just copying everything after then end of the TIFF image because you can't easily tell where the TIFF image ends.

I'll keep working on it and hopefully I can come up with a solution, but it may take a few days.

The block begins with the character sequence "CANON OPTIONAL DATA". However, this is of limited use because it would slow things down too much to search for this string.

However, I think I may have found a solution which doesn't slow things down too much. I have added some code to keep track of the last data block read from the TIFF, and simply copy anything found after this to the new file. I don't like polluting my general TIFF writer code to copy this non-TIFF information, but it seemed to be the simplest way.

I have modified the 5.64 pre-release version with this change (OK, so I was a bit quicker than I thought), and if you want to test it out you can download it from http://owl.phy.queensu.ca/~phil/exiftool/Image-ExifTool-5.64.tar.gz. I still want to do more testing to be sure this is doing what I want before I release this version officially.

Gah - spent so long getting the formatting right our posts overlapped !

Thanks for that very speedy work: I'll try it out and see if I can break it :)

--
Ian.

Excellent news ! Thanks very much for sorting it all out for me - I hadn't found any of those issues, but I was trying much simpler stuff (set an option, edit it lots in DPP, check and set another one etc.) which was mainly trying to find a point where either DPP or ExifTool got upset.

They didn't.

I look forward to the official release, but I'm off to play with this one some more now.

TTFN,
--
Ian.