Exiftool parameters used and to be used by Hugin

[Harry van der Wolf]

Hi all,

(I know that some users deliberately disable Exiftool, so this mail will not be relevant for them. Please do not use this mail to advocate for or against Exiftool use)

Currently Hugin uses the following exiftool option tags: "-ImageDescription -Make -Model -Artist -W hitePoint -Copyright -GPS:all -DateTimeOriginal -CreateDate -UserComment -ColorSpace -OwnerName -SerialNumber"

These parameters are missing the "-icc_profile" and the  "-XMP"  option tags from exiftool. I think that users working with "icc profiled" images might find strange color casts or "slightly off" colors or "faded"  colors in their final panos. colorspace is not the only parameter that should be used when using images with color profiles.

IMHO, Hugin could and should also use: "-icc_profile -XMP".  (XMP is a very detailed "array/record of info" and contains lots of exif info as well as Raw image info copied from your RAW images to jpg/tif/png etc. but also for movie container formats)
Also the option tags "-IPTC:all -JFIF:all" should be used/copied in my opinion.

I would like to add these option tags to the exiftool params in hugin (note: that is hugin wide and not only OSX).
Please let me know your arguments for not applying or absolutely applying these options to Hugin.

Also: If you think other parameters should be added or removed, respond as well.

Harry

[Markku Kolkka]

13.7.2010 10:44, Harry van der Wolf kirjoitti:
> These parameters are missing the "-icc_profile" and the "-XMP" option
> tags from exiftool. I think that users working with "icc profiled"
> images might find strange color casts or "slightly off" colors or
> "faded" colors in their final panos. colorspace is not the only
> parameter that should be used when using images with color profiles.

I agree that the -icc_profile option must be included. The documentation
for Hugin claims that the ICC profile of the image is preserved but in
my experience that's not true.
http://wiki.panotools.org/Hugin_FAQ#Why_is_the_ICC_profile_of_my_input_images_not_preserved.3F

However, I don't think that bulk copying all XMP data is sensible, it
may contain fields that aren't valid after the image is processed with
Hugin/enblend/enfuse.

--
Markku Kolkka
markku...@iki.fi

[Yuval Levy]

Hoi Harry,

On July 13, 2010 03:44:06 am Harry van der Wolf wrote:
> missing the "-icc_profile"

yes, please add.

> and the "-XMP" option

not sure about this one.

> (XMP is a very
> detailed "array/record of info" and contains lots of exif info as well as
> Raw image info copied from your RAW images to jpg/tif/png etc. but also for
> movie container formats)

XMP is eXtensible, and compensates for the problem of proprietary/special tags
in EXIF/IPTC. I have used a couple of RAW converters (both closed and open
source) that store RAW conversion parameters and other metadata in XMP
sidecars. AFAIK the data in the XMP sidecars is either redundant - copied
from EXIF/IPTC or not relevant to the image after the stitch. Moreover XMP
gets quickly bloated (e.g. with user tags, comments, ranking, etc.). I would
caution against blanket copying of XMP data. If there is a piece of XMP
information that is relevant, it can be added individually like with the EXIF
tags.

> Also the option tags "-IPTC:all -JFIF:all" should be used/copied in my
> opinion.

how much of this information is already available / redundant with the EXIF
info? I am not against enriching the output of Hugin with relevant metadata,
but the ":all" scares me...

> I would like to add these option tags to the exiftool params in hugin
> (note: that is hugin wide and not only OSX).

My 2 cents: go ahead with the ICC profile. Make a list of what is included in
XMP, IPTC:all, JFIF:all and let's have a look together as to what parts of
that information are relevant.

Yuv

[Jan Martin]

I would love to see that GPSImgDirection (the direction the middle of the pano looks) is calculated from source images, hugin settings and the added.

--
http://www.DIY-streetview.org

[Yuval Levy]

On July 13, 2010 12:47:03 pm Jan Martin wrote:
> I would love to see that GPSImgDirection (the direction the middle of the
> pano looks) is calculated from source images, hugin settings and the added.

Should not be too difficult to do: take the GPSImgDirection of a single image.
Add it to that image's yaw value and you get the GPSImgDirection relative to
the panorama's coordinates.

Yuv

[Jan Martin]

Whom volunteers to add this?
I am willing to offer a bounty of 20 $US to sweeten the coding.
Anyone to add to that?

Jan

--
http://www.DIY-streetview.org

[Bruno Postle]

On Tue 13-Jul-2010 at 09:44 +0200, Harry van der Wolf wrote:
>
>IMHO, Hugin could and should also use: "-icc_profile -XMP". (XMP is a very
>detailed "array/record of info" and contains lots of exif info as well as
>Raw image info copied from your RAW images to jpg/tif/png etc. but also for
>movie container formats)
>Also the option tags "-IPTC:all -JFIF:all" should be used/copied in my
>opinion.

I'd be suprised if there was anything interesting in JFIF metadata,
but I agree we should be copying any colour/icc stuff if we are not
already.

--
Bruno

[Bruno Postle]

There is other similar stuff that needs doing. We should be noting
useful info such as the projection type, parameters, cropping and
field of view in the comments (much like qtpfsgui/luminancehdr does
with tonemapping parameters).

More usefully, there should be enough information to allow an output
panorama created by Hugin to be used as a Hugin input photo without
having to remember what the field of view was.

--
Bruno