[argyllcms] Re: Fwd: Simple how-to on camera profiling

  • From: Pascal de Bruijn <pmjdebruijn@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 18 Mar 2010 08:30:17 +0100

On Thu, Mar 18, 2010 at 8:19 AM, Karljohan Lundin Palmerius
<karljohan.palmerius@xxxxxxxxx> wrote:
>>> I'm a "user" of colour profiling but not very knowledgeable, but I
>>> finally think that I've managed to put bits and pieces together to
>>> define a work flow for creating a settings specific colour profile for
>>> a camera. I've written a how-to on this to help others in my
>>> situation. I welcome anyone to comment on the procedure and chosen
>>> parameters. It should be noted that I use QPCard201 which has a
>>> relatively small set of colour patches and so can't be expected to
>>> suffice for a detailed colour profile.
>>> http://staffwww.itn.liu.se/~karlu/div/howto/cameraProfile.php
>> For which RAW converter are you profiling?
> I use UFRaw. Actually, I've never tried anything else but it seems to
> fit my needs.

Good. I use UFRaw and Darktable mostly...

>>> To facilitate the process of extracting the target's colour patches,
>>> I've implemented a GUI for marking the "fiducial square" to get the
>>> coordinates to put into scanin. I could not find anything like it so I
>>> just had to implement my own. It's a bit of a hack, but features
>>> rendering of the target geometry which is very useful, especially if
>>> the target is not entirely flat so that the geometry has to be
>>> tweaked.
>> Doesn't scanin autodetect this? It should...
> As far as I understand scanin is capable of finding the target as long
> as it is large on the image and only subject to affine transforms.
> This is not the case in photography.
> Also, with "not flat" I mean that the actual target can be slightly
> bent after storage in a camera bag. In that case the automatic
> matching using computer vision will have a very hard time determining
> the correct geometry even if it supports projective mappings.

Then leave the target between two books to straighten it out...

If scanin can't autodetect the target, the quality of the target shot
is highly questionable, and so will the profiling results be.

>> With low patch count targets I'd stick with matrix only profiles...
> I tried -aS and -ag as suggested by Nikolay before, but got poor
> results with -aS, so in the how-to I suggest -ag which is
> gamma+matrix. I did have some trouble also with -ag at the time I
> decided to not use -aS so the problem with -aS could have been caused
> by some other problem in the pipeline.

Again, Argyll 1.1.1 has a matrix only mode (-a m) which is very cool.

Adding some ooomph to your images should not be the job of the
profile, UFRaw has basecurves for that.

Pascal de Bruijn

Other related posts: