Feedbacks about AVM tags in Aladin

[Pierre Fernique]

Dear IVOA members,

About our AVM discussion, I spent a few hours for looking the last more 
recent press release images - concerning a sky region - that I found on 
APOD, hubble.org, Spitzer PR, Chandra PR and ESO PR (). Just for having 
a real idea of the tagging quality.

On 68 recent images, 19% have already AVM tags ! (in JPEG or just in TIFF)
But amongst these AVM tagged images, 62% have a wrong spatial reference.

               no AVM    AVM ok   AVM wrong
APOD             24         0        1
Hubble.org        7         1        2
Spitzer PR        7         1        3
Chandra PR        2         3        2
ESO PR           15         0        0

For astronomical institutes, all images have been generated by 
Photoshop. Otherwise I found several other image tools used by APOD 
users (Adobe ImageReady, Picassa, HLino?...) but in these cases without 
AVM tags.

When AVM tags are provided, the meta information seems to be correct 
(provider, bands, instruments...) but using different vocabularies 
(IR/Infrared/...).
Sometimes interesting information are not present in the dedicated AVM 
tags, but directly in the XMP envelop notably when this value is already 
defined as an XMP entry (title, description...).


My interpretation of these results.
The bad spatial references are mainly due to the 
Photoshop/Fits-liberator limitations (for cropping, rotating, resampling 
- explained in a previous mail) and as concretely it is the unique 
method for inserting AVM tags...

I am not sure that Photoshop will modify its code for supporting 
astronomical sky registration adjustements. But I do not know the detail 
of FITS-liberator plugin. Certainly their authors could have a better 
answer than me.

So especially if I'm right (Photoshop does not fix the problem), the 
main question will be:
Which dedicated astronomical tools will write AVM spatial reference 
correctly in outreach images ?
 => WCSTools ?
 => Astronomy.net ?
 => Skyview ?
 => Aladin ?
 => other ?
And amongst the authors of these tools who will agree to use XMP Adobe 
envelop ?
- by trying to use the Adobe XMP SDK toolkit ?
- by developing their own XMP writting API ?

The good (very good) point for using XMP is certainly the fact that this 
format seems to be well supported by recent image tools (Jonathan Fay's 
comments and my own tests on Windows XP, gimp2, ...). It is certainly 
the best method for having a good chance to keep the metadata even the 
image is regenerated (converted or modified & saved).

The bad point is certainly the Adobe dependence and the potential 
consequences (remember that last year Adobe Photoshop CS3 introduced a 
compact header variant for XMP tags contraining AVM to follow its 
modifications - but we can hope that this kind of evolution will not 
arrive each year)

Could I ask to the authors of registration image tools what are their 
opinion about writting spatial AVM tags using an XMP envelop ? Good or 
bad idea ? planned for your own tool ? ...

Best regards
Pierre Fernique