align_image_stack does not save/preserve colorspace exiv data?

[Harry van der Wolf]

Hi all,

I use the combination of align-image_stack and enfuse a lot for improving the dynamic range of my non-panorama photos (see article of Yuval on http://panospace.wordpress.com/2008/08/04/dynamic-range/) in case you don't know what I'm talking about. I have a lot of "church/cathedral window" pictures even though I'm not religious at all.
I sometimes run into the issue that the enfuse-merged jpg or tiff has "weird", unnatural colors. I use align_image_stack in my own droplet applications and also tried align-image_stack and enfuse on the command line to see whether my droplet apps were the problem, but they are not.
So I thought it had something to do with the source images not being correct (shit happens) or whatever, never expecting this to be a problem in the software.

However, one of the users of my droplets (expecting me to be the align_image_stack developer) mentioned to me that this unnatural (dis)coloration is caused by the fact that align-image_stack does not preserve colorspace info in the exiv data, thereby causing this unnatural coloration in the final image when enfuse merges the intermediate tiffs produced by align_image_stack.
Before entering this as a (possible) bug I would like to hear from the "color and lighting" experts in this community whether this explanation is correct and also if other users experience the same problems.

Please let me know.

hoi,
Harry

[Harry van der Wolf]

Hi again,

I mentioned " Before entering this as a (possible) bug". This is not correct and should be "feature request".

Harry

[Bruno Postle]

On Wed 13-Aug-2008 at 17:36 +0200, Harry van der Wolf wrote:
>
>I sometimes run into the issue that the enfuse-merged jpg or tiff has
>"weird", unnatural colors. I use align_image_stack in my own droplet
>applications and also tried align-image_stack and enfuse on the command line
>to see whether my droplet apps were the problem, but they are not.
>So I thought it had something to do with the source images not being correct
>(shit happens) or whatever, never expecting this to be a problem in the
>software.
>
>However, one of the users of my droplets (expecting me to be the
>align_image_stack developer) mentioned to me that this unnatural
>(dis)coloration is caused by the fact that align-image_stack does not
>preserve colorspace info in the exiv data, thereby causing this unnatural
>coloration in the final image when enfuse merges the intermediate tiffs
>produced by align_image_stack.

enfuse/enblend have a '-c' option to use icc profile information
when blending, but this is apparently very slow and you would have
to copy the metadata to the intermediate TIFF files with exiftool.

I also don't think it is likely to be your problem as the effect
should be minimal.

Have you tried just copying the profile to the output image to see
if that fixes it? ..or applying the profile to the input images
before doing anything (ie. working with sRGB throughout)?

Otherwise this could be a problem with the transparent border areas
left-over by align_image_stack, can you upload one of these weird
images so we can see it?

--
Bruno

[Harry van der Wolf]

2008/8/13 Bruno Postle <br...@postle.net>

On Wed 13-Aug-2008 at 17:36 +0200, Harry van der Wolf wrote:
>
>I sometimes run into the issue that the enfuse-merged jpg or tiff has
>"weird", unnatural colors. I use align_image_stack in my own droplet
>applications and also tried align-image_stack and enfuse on the command line
>to see whether my droplet apps were the problem, but they are not.
>So I thought it had something to do with the source images not being correct
>(shit happens) or whatever, never expecting this to be a problem in the
>software.
>
>However, one of the users of my droplets (expecting me to be the
>align_image_stack developer) mentioned to me that this unnatural
>(dis)coloration is caused by the fact that align-image_stack does not
>preserve colorspace info in the exiv data, thereby causing this unnatural
>coloration in the final image when enfuse merges the intermediate tiffs
>produced by align_image_stack.

enfuse/enblend have a '-c' option to use icc profile information
when blending, but this is apparently very slow and you would have
to copy the metadata to the intermediate TIFF files with exiftool.

I also don't think it is likely to be your problem as the effect
should be minimal.

Have you tried just copying the profile to the output image to see
if that fixes it?  ..or applying the profile to the input images
before doing anything (ie. working with sRGB throughout)?

Sorry for maybe asking a stupid question: Do you mean to use Exiftool to copy the profile of one of the intemediate tiffs into the output file?
 

Otherwise this could be a problem with the transparent border areas
left-over by align_image_stack, can you upload one of these weird
images so we can see it?

You mean that I first have to crop the images a little?
I will upload some images the images, but it doesn't happen very often and if I have "a load" of images I forget the incorrect ones. I mean that I will have to search, but I will upload them.

Thanks for your reply.

Harry
 

--
Bruno

[Bruno Postle]

On Wed 13-Aug-2008 at 20:07 +0200, Harry van der Wolf wrote:
>2008/8/13 Bruno Postle <br...@postle.net>

>
>> Have you tried just copying the profile to the output image to see
>> if that fixes it? ..or applying the profile to the input images
>> before doing anything (ie. working with sRGB throughout)?
>
>Sorry for maybe asking a stupid question: Do you mean to use Exiftool to
>copy the profile of one of the intemediate tiffs into the output file?

nona just ignores and discards any colour profiles, so you will need
to copy it from one of the input photos into the final output file.

>> Otherwise this could be a problem with the transparent border areas
>> left-over by align_image_stack, can you upload one of these weird
>> images so we can see it?
>
>You mean that I first have to crop the images a little?

Yes, there was a bug with these transparent borders, but I thought
it had been fixed.

--
Bruno

[Harry van der Wolf]

2008/8/13 Bruno Postle <br...@postle.net>

On Wed 13-Aug-2008 at 20:07 +0200, Harry van der Wolf wrote:
>2008/8/13 Bruno Postle <br...@postle.net>
>

>> Have you tried just copying the profile to the output image to see
>> if that fixes it?  ..or applying the profile to the input images
>> before doing anything (ie. working with sRGB throughout)?
>
>Sorry for maybe asking a stupid question: Do you mean to use Exiftool to
>copy the profile of one of the intemediate tiffs into the output file?

nona just ignores and discards any colour profiles, so you will need
to copy it from one of the input photos into the final output file.

Nona? Do you mean enfuse or align_image_stack. I'm not talking about an "enfused" panorama. I'm talking about a set of images to be overlayed on each other.

>> Otherwise this could be a problem with the transparent border areas
>> left-over by align_image_stack, can you upload one of these weird
>> images so we can see it?
>
>You mean that I first have to crop the images a little?

Yes, there was a bug with these transparent borders, but I thought
it had been fixed.

That brings me to another question: In svn I can see via the svn logs what has been changed. Does CVS have something equal. I have already googled for it, but couldn't find it.

--

Bruno

[Bruno Postle]

On Wed 13-Aug-2008 at 20:52 +0200, Harry van der Wolf wrote:
>2008/8/13 Bruno Postle <br...@postle.net>

>> nona just ignores and discards any colour profiles, so you will need

>> to copy it from one of the input photos into the final output file.
>
>Nona? Do you mean enfuse or align_image_stack. I'm not talking about an
>"enfused" panorama. I'm talking about a set of images to be overlayed on
>each other.

Sorry, all the hugin tools discard colour profiles (I think).

>That brings me to another question: In svn I can see via the svn logs what
>has been changed. Does CVS have something equal. I have already googled for
>it, but couldn't find it.

For enblend there is a mailing list with commit messages:

https://lists.sourceforge.net/lists/listinfo/enblend-cvs

It has a kind-of-working archive.

--
Bruno

[Daniel German]

>
> Sorry, all the hugin tools discard colour profiles (I think).
>

but panotools preserve them

--dmg

[Andi S]

So, does it just ignore the profile and re-tag it with Adobe RGB (1998) afterwards, or does it really convert colors from ProPhoto RGB (my export space) to Adobe RGB (1998) ?

The first case would be incorrect, and it should just copy the tag of the input image; the second case would be at least be correct while limiting my colors.

[Andi S]

Answer to my question: It seems that align_image_stack and enfuse pass on the original profiles and work with this data. No converting or wrong tagging is happening.