[argyllcms] Re: Number of patches vs. profile quality
- From: Nikolay Pokhilchenko <nikolay_po@xxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Sun, 08 May 2011 00:01:06 +0400
Yes, Roger's problem is likely metamerism. I have the same experience with
different printers.
My experiments with different illuminants and color management shows that with
ArgyllCMS, the prints on different printers (within gamuts) are certainly the
same if illumination spectrum is stable, measured and accounted at the profile
building sage.
If I need to build an exact profile for rather curved "RGB"-printer (HP
Photosmart 8253, for example) and need to keep dE2000<1.2, I had to print as
many as 3400-3800 patches. It's because several local non-uniformities in
printer gamut. It's almost impossible to keep Delta<1 dE2000 for any point of
gamut with certain printer. For reference: the repeatability of printer on
premium photo paper is better than 0.6dE (with 10-minutes interval between
prints). It was before Graeme implemented an optimization on preliminary
profile in targen. I haven't yet build such profiles with newer ArgyllCMS
versions.
Fri, 6 May 2011 14:57:54 +0200 Jens Heermann wrote:
> Hi Roger
>
> What makes me wonder is, the phenomenon of different visual impression while
> measuring the same Lab-values... As Lab is directly related to the visual
> perception of the human eye (CIE norm observer XYZ curves), there should be no
> visual difference. What is found quite often, is that identical patches (in
> terms of their reference values) are perceived differently on two printouts,
> but when measured, these patches will also create different Lab-values.
> Getting the "same" Lab-measurements on different printers confused me also...
> Due to a normal measurement deviation you should get different values in a
> range of roundabout dE0,5, depending on the measurement device, even in
> measuring the same patch on the same substrate more than one time.
> No offense, but maybe you are a little imprecise in describing the problem and
> your "workflow", which makes it hard to analyse the cause.
>
> So, I assume you are able to create standard-conformant printouts, but the
> visual impression on different printers is also different, even if the average
> dE is near to same on both printers. Is that correct?
>
> What is the lighting condition under which you evaluate the printouts? Maybe
> this phenomenon is caused by different reactions of the different colorant in
> the different inks to the lighting condition (aka metamerism).
>
>
>
> On the general topic my option is, that the more patches you measure the more
> "support points" you get for the profile creation the more precise your device
> is characterized. Also measuring more patches means more time needed, so you
> have to find a practical compromise between the number of patches and the time
> you like to invest into your profile quality.
> The other thing is that, some devices (and/or the combination of printer and
> measurement device) do not react completely linear, so you may get in some
> areas inhomogeneous measuring data (lots of positive/negative peaks) which may
> lead to tearing in gradients, because the behavior is documented to precisely.
> For the optimal visual output it may be better, if more interpolation in those
> areas is used to smooth the profile data. Especially the i1 pro is quite
> unstable with dark patches...
>
> Regards
> Jens
>
>
> -----Ursprüngliche Nachricht-----
> Von: argyllcms-bounce@xxxxxxxxxxxxx [mailto:argyllcms-bounce@xxxxxxxxxxxxx] Im
> Auftrag von Roger Breton
> Gesendet: Freitag, 6. Mai 2011 14:02
> An: argyllcms@xxxxxxxxxxxxx
> Betreff: [argyllcms] Re: Number of patches vs. profile quality
>
> Hi Nikolay,
>
> I am curious to study the role of colorants in a CMYK printer (could be
> applicable to RGB too). But the problem I run into, profiling CMYK printers,
> is that, using a constant substrate, depending on the printer model (Epson vs
> HP), I get very different visual appearance for the same CIE Lab measurements.
> I don't know where to start the study of this "problem", because it is a
> problem for me and I'm sure many other users. Is this something I can see in
> the SPDs of the "solids", the CMYK primaries and their "overprints", RGB? Or
> in the way that the grays are formed? Colorimetrically, these printers will
> create perfect match to any reference conditions. Very close to 0.35, 0.5
> deltaE. But visually, it's garbage. So reddish, for instance. I won't mention
> printer model. And this is an effect I observe all the time.
>
> I wish I would know where to start...
>
> Best / Roger
>
> -----Original Message-----
> From: argyllcms-bounce@xxxxxxxxxxxxx [mailto:argyllcms-bounce@xxxxxxxxxxxxx]
> On Behalf Of Nikolay Pokhilchenko
> Sent: May-06-11 7:39 AM
> To: argyllcms@xxxxxxxxxxxxx
> Subject: [argyllcms] Re: Number of patches vs. profile quality
>
> Hello Graeme! Hello list!
>
> It's interesting question.
> From my experience, there is not enough 1400 patches for profiling "unknown"
> RGB printer (Canon or Epson with non-original inks). The most problems are in
> the darks: the printer and paper combination may have an unpredictable
> behavior (steepness of device response in dark region) and the 1440 patches of
> OFPS is insufficient. With 1440 patches, darks are not well characterized in
> 50% of cases. There may be few patches per volume unit of the darks.
>
> My stable workflow for distant profiling of "RGB" inkjets is 2880 patches on 2
> pieces of A4 sheets (6x6mm patches). The 2880 patches is enough to profiling
> any printer in good condition. In some cases one of 2 sheets is "broken" - by
> customer error or for another reason. In such cases I do the profiling only by
> 1 sheet. It's mostly OK.
> There is critical patch R=G=B=0 and if it's absent or misreaded, the profile
> may become incorrect in darks - there may be noticeable hue shift in blacks.
> So I ask to add more black patches to RGB chart - as many as white patches. As
> I understood, the only one black patch is important for determination of
> device black point and it's color value is critical. So it will be more
> correct to have several black patches on the chart by default.
>
> When I profiling wide format CMYK inkjets, usually the profile with 1200..1400
> patches is OK. But the device is linearized previously - by intrinsic RIP or
> by printcal.
>
>
> Wed, 04 May 2011 13:11:45 +1000 Graeme Gill wrote:
>
> >
> > I am interested in hearing about peoples experience with profile
> > quality vs. number of test patches using Argyll. In particular, I'd
> > like to get an impression of what aspect of a profiles quality are
> > seen as improving as more patches are used. If this aspect is color
> > accuracy, is there a particular area of the color gamut that is seen
> > as the critical area ? If so, what area is it ?
> >
> > [I'm wondering if the profiling efficiency can be improved by some
> > simple changes to the patch distribution.]
> >
> > thanks,
> >
> > Graeme Gill.
Other related posts:
