The peak DeltaE is not more than 3.0 under G7 or ISO

[terryhucoco]

WHU colorproof v2.0 can do this with no circle caculation .Actually,the peak DeltaE is 2.33 (CIE76)under ISO,do you think so? The pictures can tell you the truth:

 

[rich apollo]

What is meant by "circle calculation"?

 

[terryhucoco]

cicrle caculation always be used by software to decrease DeltaE ,such as mx4(used by GMG) is created by cicrle caculation.

 

[meddington]

I'm taking it to mean iterative profiling.

 

[terryhucoco]

Yes, and the purpose is to get better data.But how about the millions of CMYK composition unrecorded in the it8 target.

 

[meddington]

But how about the millions of CMYK composition unrecorded in the it8 target.

I've tested this years back to satisfy my curiosity with GMG's iterative profiling. Comparing a 7000+ patch color target before and after profile iteration that used a smaller 1200 patch target. There was clearly improvement in the patches that weren't present in the smaller target. The question remaining was could that level of accuracy be present in a profiling process that offered no iterative step. This was years back though, when there was more of a difference in the level of quality one could expect from profiling solutions.

 

[terryhucoco]

I think iterative profiling is not the right solution to solve digital proof problems.the data of the 7000+ patch color target is caculated by GMG,and of course,it is not the actual data of the press.the scource data is not right,how about iterative profiling ?So I think if we can get the most accuracy data in 1200+ patch target at one time,the millions of color can be under control.

 

[Lukas Engqvist]

@terryhucoco The IT8 doesn't have millions of colours. And I don't quite understand the purpose of the thread. It is meerly a statement? Or are you asking for comments on your statement?

 

[meddington]

the data of the 7000+ patch color target is caculated by GMG,and of course,it is not the actual data of the press..

In this case it actually was. I didn't explain fully, but in this case, both the 7000 patch target and 1200 patch target were printed off a Kodak Approval...the device used as the source to be simulated on an inkjet proofer. The smaller target to establish the source data and used in the iterative process. The larger target used merely as confirmation before and after the profiling process, but not used in any calculation. This showed me that the iterative process using a smaller patch target did indeed lead to homogenous improvement of colors outside the smaller patch set. This is as one would expect with any profiling process of course, but at the time the was skepticism that GMG was doing anything more than optimizing the smaller patch set, which obviously isn't the case.

 

[Davis.zhou]

iterative profiling is a scientific sampling method not only use in getting a better icc or proof.
X-Rite use this in colormunki for monitor profiling, in IPEX 2010, X-Rite announced the i1-Profiler will have this in Q4-2010. Xerox also file a patent of their own iterative profiling in 2009 - Control base on iterative profiling methods (Pub. no: 20090296171). If you search "iterative profiling" by Google, you will find that "iterative profiling" also applied in IT, project management, bio-chemistry, etc.
The basic idea of the iterative profiling is using the less samples in a resonable time to get a close enough and accurate result instead of measuring all data. Sometimes, too much data for measuring will cause error by measuring deviation.

 

[Al Ferrari]

"But how about the millions of CMYK composition unrecorded in the it8 target"

The way I read it, by referring to the colors NOT in the target, he too is saying that the IT8 doesn't have millions of colors. But I am not sure about the purpose of the thread either. Perhaps Meddington can shed light.

Al

 

[Lukas Engqvist]

Al ty missed the "un-" oops.

 

[meddington]

. But I am not sure about the purpose of the thread either. Perhaps Meddington can shed light.

Not positive, but I'm guessing the thread started as an unsolicited promo from a vendor.

 

[Johu]

Just by quickly looking at Terryhucocos attachment I could see that there are bigger peak values than 4.7. You will see always smaller DeltaE values when you use DE2000, which is the formula used in the attachment (also c´learly stated in the header). Basic DeltaE (76) gives you higher peak values. Also you get lower average values if you use the "old" two row wedge and not 3.X. Nevertheless it is not bad result, but can not compete with iterated results. Also I should make a point that measured values are not the only guarantee of quality. One should always check other visual things, like smoothnes of gradients etc.

 

[terryhucoco]

Yes,the peak deltaE is 2.33 according to CIE76.I can not use the three row wedge because of copyrights.So I give the free Gracol control wedge.If you need ,I will give the result of it8 target more than 1000+ patch.Of course,the peak deltaE is not more than 3.0 without using iterative profiling .My propose is that if we use iterative profiling ,we can not solve problems of different cmyk compositions.Our target is only 1700+ patch and we can not promise that the millions of color are all under control. if the press color is c95m85y45k75(not in target),we cannot get the accurate color from digital proof.

 

[Lukas Engqvist]

So what you are in fact saying is that even though you get "good" delta-E values for the patches you have, you feel that this is inadequate to prove that the proof is correct enough for all colour representation. That those colours not on patch may be way out, and that iterative processes will not come to terms with this problem. (Sorry for me the non-english table and that I have not seen a report exactly like it putts me off, and i realise this is a weakness on my part, but makes it hard for me to comment on the various parts of the report.)

Is the printer infact a CMYK printer? If you are using an ink set or technology that is different to the one you are proofing, yes the question is justified. But this is why the patches that are chosen are key. If you compare a standard offset press with an Indigo, you will find that the patches are designed to see primaries, secondaries and tertiariy colours. When technology differs, you may have very similar primaries yet the secondaries will be very off due to trapping and transparency of the inks being different. The manufacturers will then maybe compromise the primary pigments so that totally you will get better primary and secondary colours. In theory you could have primary colours wich had a large delta-E and secondaries that have a larger delta-E than either of it's primaries, but usually the manufacturers strive for the opposite. This is why in the report you have the graphical charts, so that you would be able to identify inconsistencies.
On a 6, 8 or 10 colour printer this becomes more of an issue since there are more levels of mixing. I do not have a full enough understanding of those colour models to say how many patches would be required, but I assume the G7 patches would fall short of giving a representation of the gamut and how the secondaries, tertiaries (how many colours can interact in those systems? I mean strictly speaking an epson-6 colour proofer with CcMmYK would be secondaries on an Cyan or Magenta wedge)

But even if you look at monitors they are often just profiled with a few patches. To profile a monitor with more than 99 patches would not only be overkill, but may also introduce error n much the same way as a TVI curve will not benefit from having too many control points.

In short what I am trying to say is that as with much science, by accurately measuring the extremities we can extrapolate with reasonable accuracy the in betweens. If looking at grey scale, you can with 7 standard patches generate a TVI curve, and then you will find that if those patches are within tolerance the grayscale images will be accurate wether 8, 32 or 64 bit, just from calibrating the 7 patches. (and possibly disgarding 3 patches for a smoother curve will give you even better accuracy )

 

[martin.weberg]

But even if you look at monitors they are often just profiled with a few patches. To profile a monitor with more than 99 patches would not only be overkill, but may also introduce error n much the same way as a TVI curve will not benefit from having too many control points.

Hi Lucas

The problem does not lie in how many samples you've got. Really, you can't get too many (there are however practical/economical issues in using too many samples, this is where statistics come to play). The trick is how you model the device from measured samples (using smoothing algorithms).

Martin Weberg

 

[martin.weberg]

In short what I am trying to say is that as with much science, by accurately measuring the extremities we can extrapolate with reasonable accuracy the in betweens.

Hi Lucas

Sorry I'm picky today ;-)

But extrapolating is always sketchy, since you know nothing about the device/process/model behavior beyond the sampled range. For sure you mean interpolation, right?

Martin Weberg

 

[Bloodsaler]

During the Iterative profiling,lower Delta E value doesn`t mean correct Color.From my experiance with EFI Colorproof,I don`t see any significant improvement during Iterative profile process.Otherwise,to many times iterative profiling may cause some tone scale problem.
But actually the DeltaE value decreased in system.

 

[Lukas Engqvist]

@Martin, thank you