[argyllcms] Re: CMP Digital Target 3 (not 003)

  • From: Pascal de Bruijn <pmjdebruijn@xxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Sun, 8 Mar 2009 14:06:01 +0100

2009/3/7 Nikolay Pokhilchenko <nikolay_po@xxxxxxx>:
> Pascal de Bruijn wrote:
>
>> I'd like to hear your opinion on the target designs, and what is expected
>> to have the best results with Argyll.
> ...
>> I'm aware it's hard to comment on production quality... But assuming
>> production is up to snuff... Which target is "best" designed? Or shouldn't
>> I expect any better results than with my IT8 target?
>
> I think, the more patches on target will give the more accurate results. But 
> when the angle size of a patch is rather small for your camera, the lens 
> flare can increase measurement error. I intuitively prefer the chaotic patch 
> distribution on target, but it worth to be mentioned, that depending on lens 
> quality and shotting conditions, the patches can influence each other.
>
>> Does anybody here have ColorCheckerDC reference files? I can't
>> find them on X-Rite's website.
>> Are ColorCheckerDC targets individually measured? or batched?
>> or are they manufactured to a very small standard deviation?
>
> I have ColorChecker SG. You can't find reference on X-Rite site, because the 
> Color Checker is intended to use with X-Rite software only, I suppose. I 
> found one file in X-Rite folders on my computer. But I don't use them. I 
> built my own.
> I've attached spectrum measurements of my Digital ColorChecker SG in 
> DigitalColorChecker_Nikolay_Po.zip. The measurements was 3-time sampled and 
> averaged. I've included corrected Graeme *.cht (with slightly more wider 
> borders between patches because the shadow from plastic frame presence) and 
> D50.cie and D65.cie - XYZ/LAB refrence for my particular target copy in 
> addition. I think, there is very small standard deviation in production, 
> because I've get no any reference file in shipment.
> If You want to compare different targets, I recommend you to see the spectra 
> data.
> I think, the 140 patches on DCCSG is rater small, so, I can't recommend you 
> to buy smaller DC ColorChecker. IMHO.
>
> By my experience, with ColorChecker SG only the matrix profiles can be build 
> for my camera with 140 patches of ColorChecker SG. The LUT profiles haves 
> deviations (see below) and in generally don't applicable even for simple 
> daylight illumination. Furthermore, only the matrix+gamma profiles can 
> adequate (but not exact) predict the out-of-target-gamut colors, while the 
> matrix+shaper profiles are often incorrect, the LUT profiles has huge 
> unpredictable errors. I wasn't able to take even one successful image with 
> LUT-profile, but successfully taken images with matrix profiles with Argyll 
> CMS (I haven't use other software). It seems that single gamma profiles 
> generally better than different channel gamma.

So far, with my Wolf Faust IT8 target, I've only been also to build
proper matrix profiles as well. XYZ LUT profiles never really worked
for me either. The IT8 target has 288 patches.

Though I've been building Matrix + Single Shaper, not Single Gamma. A
gamma curve seems too basic to accurately portrait my camera's
response curve. I have hacked colprof a bit, to get extra smooth
shapers.

I've tried Multiple Shapers too, but got odd (and horrid) results. It
seems logical that a single shaper should be used, as there is no
direct reason to believe that the CMOS/CCD sensor should have a
different response curve for each of its color components.

> I've bought my first target from Ukraine for $32 in 2006. They have 288 
> patches printed on multiink inkjet printer with high gamut.

I guess this was IT8 as well? Kodak Q60 perhaps?

But did you get better results with a DCCSG than your own IT8?

Regards,
Pascal de Bruijn

Other related posts: