Any IIIF servers retain an embedded ICC color profiles in derivatives?

[Beaudet, David]

Does anyone know which, if any, of the IIIF image servers, regardless of commercial vs. open source status, retains the ICC profile in derivatives provided one is embedded in the original?

 

In other words, if there's an embedded ICC profile in a JP2 source image and the following ImageMagick commands are run:

 

convert source_image.jp2 source_image.icc

and

convert iiif_derivative.jp2 iiif_derivative.icc

 

will source_image.icc exactly equal iiif_derivative.icc ?

 

Thanks,

 

Dave Beaudet

National Gallery of Art

[coms...@g.harvard.edu]

Hi David.

It is my understanding that the Loris image server correctly embeds ICC profiles in the JPEGs it spins out for web-delivery from JP2 source files. This is based on my recollection of conversations from a couple of years ago, but I believe that Princeton's Jon Stroop upgraded Loris to generate correctly tagged JPEG files for delivery.

I also remember (think I remember) from those same conversations that the IIP Image Server added ICC support to their development roadmap, and I would hope that by now that software too incorporates ICC color profile tagging. Maybe someone else on this list knows the status of ICC support within IIP or within other image servers used to generate JPEGs on-demand from JP2 source files.

Best,

-- 

Bill Comstock
Head of Imaging Services
Harvard Library Preservation
Widener, D70C
Harvard Yard
Cambridge, MA 02138
--
P 617-496-5241
F 617-495-0403
http://imaging.harvard.edu/
--
* Digitized collections at Harvard *
http://library.harvard.edu/digital-collections

[Jason Ronallo]

Does anyone have a JP2 they could share that has a known good embedded
ICC profile that I could use to test whether it would be possible for
me to add support to the Node-based IIIF image library that I wrote? I
think the binaries and libraries I use support it, but have never
checked.

Jason

> --
> -- You received this message because you are subscribed to the IIIF-Discuss
> Google group. To post to this group, send email to
> iiif-d...@googlegroups.com. To unsubscribe from this group, send email to
> iiif-discuss...@googlegroups.com. For more options, visit this
> group at https://groups.google.com/d/forum/iiif-discuss?hl=en
> ---
> You received this message because you are subscribed to the Google Groups
> "IIIF Discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to iiif-discuss...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Jon Stroop]

Jason,

https://github.com/loris-imageserver/loris/blob/development/tests/img/47102787.jp2

This file was supplied by Bill.

The labels of colors won't line up with the colors in which they are
rendered if the color profile is not applied, e.g.:

Correct:
http://libimages.princeton.edu/loris2/47102787.jp2/full/full/0/default.jpg

and an incorrect rendering is attached.

-Js

[David Beaudet]

This is very helpful.  Thanks!

I have a follow-up question.  Will Loris (and other IIIF servers) work with a JP2 color depth (bits per pixel) of 16?  The JPEG (non-JP2) only supports 8 bit color, so curious if Loris and other servers will automatically reduce the depth to 8 bits.  PNG seems to support 16 bits per pixel.

[Ruven]

IIPImage handles all standard bit depths, including 16 bit per pixel per channel images. For conversion to JPEG, the data is, by default, linearly scaled to 8 bits. If you are using TIFF, it's also possible to specify min and max sample value TIFF tags to get a different offset and scaling factor for conversion to 8 bit.

IIPImage also has (non-IIIF) commands to allow the user to specify a custom gamma and scale factor if you want the user to be able to define a custom non-linear scaling to 8 bit.

Ruven

[Lukas Rosenthaler]

Have a look at http://sipi.io . It preserves all metadata including ICC profiles and goes a hard way to preserve event profiles that are not allowed in JPEG2000.

Unfortunately it requires the kakadu JPEG2000 library which requires a license. The rest of the code is open source.