CHROMiX

VPR curves spectral data, but can't curve RGB spectral data?

Well this seems to be strange, VPR on Curve 3 advertised to be be able to curve spectral data, so if I have RGB target measured and converted to CMYK + spectral data the Curve 3 crashes.

Spectral data is just that spectral data, if Curve 3 can curve it then I see no reason why it should not work. If only CMYK targets can be curved then perhaps Curve 3 should not be advertised as working with every print process.

What Curve3 does in VPR is start with a CMYK P2P target, take the curves that it suggests and apply them to a profiling target. Most people using Curve3 have a CMYK RIP so I am not sure why you would want to curve RGB data. Are you wanting to make an RGB profile while printing on a system that uses a CMYK RIP? Tell me more information about how you can use this and we can see if we can implement it. So far I think this is the only time I have heard of someone asking to apply VPR to RGB data. It would be great to hear how would you ultimately use it if you had this feature.

It is possible to use the Photoshop curve or the Device Link profile in the Curve3 export in order to curve, using Photoshop, an image from any print system. That would be a way to curve an RGB image using Curve3 data, and it does not make use of a RIP. Let me know if you need more information on that.

I have not seen the Curve3 crashing that you describe. When I attempt to use RGB file in VPR, the software pops up an error messages telling me that the file must be in CMYK format for VPR function

Well it seemed to me that if curve3 curves spectral data then any spectral data should be curved. I’m thinking the limitation is that it must find CMYK patches for curves right? And on RGB data there may not be a match etc.???

Well I want to use it on a RGB printer instead of RIP. Like in documentation I want to make Cuve data + ICC profile. Then instead of re-creating ICC profile wasting many sheets of paper (paper can be very expensive) I would just print P2P target (way smaller than any ICC target) and curve it to match my profile.

This would save me and many others too: time, money, paper, inks etc.

The ability to have “linearization” feature on RGB printer would be enormous, CMYK native printers still speak RGB in most cases so they would also benefit if user does not want to get a RIP.

Yes I have exported the PS curve profiles and making tests like this:

original ICC profile created for paper
Curve (various settings) to get optimal curves for device (so far feather off 0-100 works best)
Curve the reference target with optimal curves
Compare the ICC original profile without curve3 and ICC profile with curve 3 curves applied to it.

Will post results.

I tried to convert the RGB + spectral data converted to CMYK by patchtool to curve 3. I really wanted it to work !

The prints after using curve + profiling software are worse than without using curve3. Gray balance is bad.

The curve3 only prints themselves are very good, even comparable to ICC profile (I mean gray balance comparison). But if both used together than that obviously doesn’t work.

Is this because curve3 can’t be used with RGB printers, will they be supported?

I’m sure many would like to use VPR to average their RGB charts too.

I was expecting that the resulting ICC profile quality at least would be the same, but it seems either the profiling software does not like curve 3 or vice versa. In theory there should be no problems, better grayscale before profile is made the better, but somehow this ain’t true.

I sent you PM with measurement data.

Curve3 is a great linearization tool to linearize printers using the G7 method. If you are printing through a printer driver, then you will already have a good linearization in the media settings that you choose. All you would need to do is print a profiling target image and measure that to get a good profile of your printer. I dont know of a method to take a P2P target to curve it to match a profile. Depending on the options you choose, it can take several sheets of paper to make your profile, but once you have made a good profile, should not need to make it again unless something changes with the printer, or the ink or the paper.

I understand there is sometimes a desire to do a separate linearization of an RGB process.
I guess Im not clear at all on the steps you are taking to do what youre trying to do. You sent me measurements of a control strip: the original measurement, the measurement after Curve3 curves, and with Curve3 and ICC profile. I dont know what steps you took to create the last measurement. I would also need to see the profile youre working with.

Page 72 in the manual talks about using the DLP export in order to curve images in Photoshop. You would want to follow those steps very precisely in order to apply the color transformations properly. This explains how to apply the curves in an RGB process, and youre right - the results are very good, even without printing through the usual device profiling transform. I dont know of anyone who has tried to do this DLP transform and then tried to follow it with a normal color management workflow. It sounds like you might end up double-profling if you tried that.

Page 56 in the manual talks about using Curve3 to smooth measurement data. This would be of benefit to RGB profile makers.

I don’t use manucfacturers OEM ink. So the canned linearization curves are not correct. Also If I would use OEM ink but not OEM paper then the linearization curves are also not correct. The RIP for small format like ZEDONET PrintFab would benefit from curve 3 also if it would work with afterwards created ICC profile.

I don’t understand what you wrote here. As I understand curve 3 linearizes printer for G7 grayscale method. Since better Grayscale means better profile created afterwards then why from the data I sent you this clearly is not the case.

Both standalone programs the ICC profile and G7 Curve 3 makes good grayscale but together they do not work. But this is the idea in the Curve 3 quotes form the manual:

“As Printed leads to smoother calibration curves and better ICC profiles, with little change in overall appearance.”

“This can result in much smoother tonality in high density areas of the image, and lead to better ICC profiles.”

“If you want the Custom White point to remain visible with an ICC profile, print the characterization target through curves made with the default Paper White setting, then after creating the profile, re-calibrate the printer with the desired Custom White setting.”

“Correction curves calculated from the P2P target are applied to the characterization data set to produce a new data set similar to what
would have resulted if the target was re-printed through the G7 correction curves. From the new characterization data, an ICC profile can be produced representing the device in its optimum calibrated state.”

I can send you ICC profile. The steps were:

“the original measurement” - was made with ICC profile only representing how good the ICC profile works.
“the measurement after Curve3 curves” - was made with Curve 3 curves photoshop, representing how good the curves works.
“with Curve3 and ICC profile” - was made with curve 3 curves and ICC profile together. The workflow was:

  1. A test target that I sent you was opened in Photoshop, Curve 3 curves applied.
  2. The target was printed using ICC profile that was created with Curve 3 curves applied to reference.

Please read above Quotes from your manual.

[/quote]

I tried it does not provide better data.

I sent you PM with link to the profiles

Don Hutcheson has recently written about the steps involved in using Curve3 with Photoshop. I will reproduce his suggestions here. This will describe more precisely what steps should be done to do this using Photoshop. This is the only way I have heard of to make use of G7 correctly in an RGB workflow. If you are trying something different, I don’t know if it will produce good color or not:

My workflow was exact like you say a post above in 14 steps.
The curves in photoshop was used instead of DLP, and target was i1Profiler RGB target, not CMYK.

I figured I rather have software convert measured LAB data to CMYK to use in curve 3, and not use CMYK target to be converted to RGB by printer and not have any control of the conversions. Was that a mistake that this made worse results with curve 3 curves than ICC profile alone?

I can make the test again using like IT8.7-4 target.

Working in Lab in Photoshop should be fine.

I did not use LAB in Photoshop, I used the RGB target like I said.
The P2P target is CMYK and my printer is RGB controlled not CMYK.

I used Photoshop to convert the P2P target to Adobe RGB by using ICC profile workflow. I printed the target by using “no color management” WINXPSP3 by using Qimage app.

The target was measured with Isis, and then data was loaded into Curve 3.
Curves were calculated and saved as photoshop curve file.

The ICC profile target was curved in Photoshop using this acv file and printed same way as P2P target.

The ICC profile was generated and used on the test target used to see if Curve 3 helps to get better ICC Profile than without it. Result was as you know not a positive one.

Do you see any mistakes in my workflow, perhaps I should use DLP profile.

The read me file for P2p target notes that:

“Note that the official supplied TIFF target images use LSB (least
significant bit) dithering to simulate better than 8-bit accuracy. This extra
precision will be lost in custom-generated targets.”

“Custom-generated target images MUST contain fractional percentage
values, especially in light patches below 10%. Rounding fractional
percentages to integers may lead to inaccurate G7 calibration.”

Do you have P2P target as RGB or LAB target? Maybe the problem is the conversion from CMYK to RGB? I’m sure that P2P started as LAB reference values not as CMYK values as now Curve uses spectral data.

Do you have P2P target as RGB or LAB target? Maybe the problem is the conversion from CMYK to RGB? I’m sure that P2P started as LAB reference values not as CMYK values as now Curve uses spectral data.