Re: QPcard 201 with gnome-color-manager

[Pascal de Bruijn <pmjdebruijn pcode nl>]

On Fri, May 4, 2012 at 11:37 AM, Richard Hughes <hughsient gmail com> wrote:
> On 4 May 2012 09:58, Pascal de Bruijn <pmjdebruijn pcode nl> wrote:
>> Just curious for which camera? And for which application?
>
> Canon something or other. Sorry.

I'm asking, because with Darktable in particular we ship custom
profiles already for many camera's.

If the camera in question doesn't have a custom profile yet, I'd be
interested in getting target RAWs :D

> We wanted to see what the profiles
> would look like with different illuminates compared. Instead we got
> the profile attached which is just batshit-insane.

Yes, I see.

>>> I know the
>>> limited number of patches aren't going to produce a super profile
>>
>> The ColorChecker doesn't have many patches either, but it's fine for
>> matrix profiles. I'd would probably use colprof -aG for something like
>> this.
>
> Yes, agreed. We've got this in the code already:
>
>        /* check whether the target is a low patch count target and generate
> low quality single shaper profile */
>        if (reference_kind == GCM_CALIBRATE_REFERENCE_KIND_COLOR_CHECKER ||
>            reference_kind == GCM_CALIBRATE_REFERENCE_KIND_QPCARD_201)
>                g_ptr_array_add (array, g_strdup ("-aG"));

I probably insisted on this some time ago :D

>> I think you need the .cie file.
>
> Great, I've added code in GCM to not ask the user for this reference
> CIE file and just use the one from argyllcms.

Right. The same thing would be possible for (real) ColorCheckers.

Though for example I print and measure my own DIY ColorCheckers (
http://blog.pcode.nl/2010/02/13/homebrew-colorchecker/ ), and those
obviously won't work with argyll supplied reference values.

Then again, I usually use the cmdline tools for camera profiling. I've
grown used to them.

>> I'm guessing the procedure with the QPcard is similar to that of the
>> ColorChecker, in that all the card are produced to a tight
>> specification to match a reference.
>
> Right, that sounds about right from what mizmo said.
>
>> One of the thing that may be a problem is gamma... You can pass -G 1
>> to scanin, especially if the RAW converter output is still linear.
>
> Is there any way to know this automatically? Thanks dude.

I'm not sure, I don't know if this is encoded into EXIF, and even if
so, I'm not sure this is reliable.

The thing is that it really matter how you expose a chart for the RAW
processor in question. You need to configure the raw processor the do
RGB pass-through. Which means you need the RAW processor's input color
profile to be the exact same as the output profile (usually applied
with a absolute colorimetric rendering intent).

And then for example UFRaw has gamma/linearity settings, you need to
set to 1.0 (which defaults to 0.45, which is 2.2 inverse I think). In
Darktable's case, we have vendor specific basecurves for that, which
you may or not may not to keep enabled.

Also exposure can matter at lot, you basically want to check the
brightness of the most white patch and see if it matches the card
reference values in the .cie file. If there is still a large
discrepancy, re-shoot the target. If there is a small (+/- 0.3EV)
discrepancy digital exposure correction may be applied.

For darktable in particular I wrote (and screencasted about this):

http://blog.pcode.nl/2010/06/28/darktable-camera-color-profiling/

Regards,
Pascal de Bruijn