Scanner Profiling with it8 Target

[nirpat89]

I was trying to learn a little bit more about scanner calibration issues and I noticed that the data file accompanying the it8.7/2 reflective target that I have specifies the illuminating source at 5000K.  My monitor is profiled at 6500K.  I have no idea what the color temperature of the scanner lamp is.  To get the best out of this calibration process, should they all have matching color temperatures? 

[Mark D Segal]

The profile describes the colour rendition behaviour of the scanner. When you make the scanner profile it records differences between the scanner target's reference values and the values the scanner returns when it scans that target. This is independent of anything you see and has nothing to do with your monitor settings. I wouldn't be concerned about the relationship of the scanner lamp to the 5000K specification for the target. Nothing you can do about it whatever the lamp temperature is. The main thing is that the profiling process describes how the scanner sees the colours of those patches and usually it does exactly that. Where your monitor comes in to the picture, literally, is what the scans of real photographs look like on your display. Colour appearance at 6500 will be somewhat cooler than colour appearance at 5000. How you set your display parameters should best reflect what works best for you when you are softproofing for printing.

[nirpat89]

Mark: 

Thanks for the clarification. 

I intend to scan some of my POP contact-prints so I can share them on the web.  For that I wanted to come as close as possible to reproduce the tone of the real print when displayed on the monitor.  I noticed that the if I let the scanner software (Vuescan) apply the profile generated from the it8 target, the resulting scan was more deviated (strong green cast) from the actual print than the "uncorrected" default scan.  Hence the question as to the source of this discrepancy.   

:Niranjan.

[Mark D Segal]

To start with, if you are sharing stuff on the web, colour management is what we call a "crap-shoot", because everyone who looks at them will likely be doing so with different monitors set in different ways. Ever gone into a TV store where they have a pack of TVs all displaying the same channel and every picture looks different? OK you understand. So beyond what you post, you do lose control over what people actually see.

From what you are saying about your Vuescan experience, it looks as if the profile you are using is no good. Perhaps generate a new one. I don't know what profiling devices and software you are using so can't advise. Tell us a bit about your profiling approach, equipment, software, settings, etc.

[nirpat89]

To start with, if you are sharing stuff on the web, colour management is what we call a "crap-shoot", because everyone who looks at them will likely be doing so with different monitors set in different ways. Ever gone into a TV store where they have a pack of TVs all displaying the same channel and every picture looks different? OK you understand. So beyond what you post, you do lose control over what people actually see.

From what you are saying about your Vuescan experience, it looks as if the profile you are using is no good. Perhaps generate a new one. I don't know what profiling devices and software you are using so can't advise. Tell us a bit about your profiling approach, equipment, software, settings, etc.

Yep, I am aware of the "crap shoot."  That's why I just want to do a good enough job and not get too bogged down.  Since these are monochromes, they will be easier to deal with if there a color cast.  I already had the standard target, I thought I will use it and see if I get a better outcome (which it does not seem to be.)

Here are the particulars:

Epson Perfection 3200
VueScan (9.5.75)
it8.7/2 target on Fuji Crystal " Procured from Wolf Faust

VueScan has a "Profile Scanner" task that will get you scanner icc profile based on scanned target and specified data file containing the readings supplied with the target. 

You can use this profile then in the color settings for any new scan.

Here are scans of the target with (-0001) and without (-0002) the scanner icc profile applied.  You can see the greenish cast in the former by looking at the border outside of the target (which was a black cardboard placed on top of the target while scanning.)  I did repeat the profile generation a couple of times more with no discernible difference.

:Niranjan.

[Mark D Segal]

The target should be good. From what you say here, looks to me like either a Vuescan problem generating the profile, or an issue with the scanner hardware.

[Mark D Segal]

By the way, looking at those two images of the target you posted, neither of them are correct.

[nirpat89]

By the way, looking at those two images of the target you posted, neither of them are correct.

It looks almost like the average of the two would look about right.  One is too magenta, the other too green.

[Mark D Segal]

Correct.

[nirpat89]

Just for the heck of it, I blended the two 50/50 and here is the result.  It looks much more closer to the target hard copy on my monitor, which I am pretty sure is not close to being as accurate as would be yours.

[Mark D Segal]

It looks like an improvement, but there is an apparent colour inconstancy from patch to patch in the grayscale (or what is supposed to be a grayscale) underneath the colour patches.

[BradSmith]

I defer to Mark's far superior expertise than mine, but when I read the RGB values of the horizontal grayscale bar in your scan 0001 on my monitor using the application "Digital Color Meter" on my Mac, I get mostly pure gray readings with R, G and B being equal.  Once you get to the darker values, it starts to wander some toward blue, depending on where in the tone rectangle I'm reading. For whatever it is worth, on my monitor, the pure white in lower left reads 218, 218, 218 and the black in darkest patch in the lower right reads 24,24,24.    And the vertical bar, #16 values all read as pure gray as well.

And your 0002 scan has a decided red cast visually (not green) and the Digital Color Meter values show a deficiency in R. 

I'm on a Spectraview calibrated NEC 2690 on a Mac, reading your files in Safari.

Brad

[Mark D Segal]

Hi Brad,

Actually I was relying on my eyes when I made that comment. Many of those patches look off-neutral to me. My display is an NEC 271PW profiled with BasicColor, so what I see is as reliable as my eyes are - if they are, my annual check-up isn't quite due yet  :-). The most reliable thing to do would be to examine the profiles themselves in an application designed for that purpose.

[JayWPage]

Niranjan,

It's not entirely clear to me what you are doing, but let me speculate: you have tried to calibrate your scanner using an it8.7 target and Vuescan, then have scanned the target again [and applied the calibration] to produce an image.

There are quite a few things that can go astray while doing this. The calibration text file must be present and it's location properly identified during the calibration scan, the file generated by the calibration (scanner.icc) has to be saved in a location where it will later be accessed by Vuescan during scans requiring calibration adjustment. Those later scans must have the input->media field set to the proper value for the calibration to be applied.

In addition, when scanning to produce an image you have to chose the correct white balance (regardless of whether the calibration is applied or not). This looks like a white balance problem to me. I find in Vuescan that manual selecting of the white balance works best (right click on a neutral grey patch, RGB values equal and about 120).

[nirpat89]

Thanks, Brad.  I appreciate your input. 

I wonder if what Mark is alluding, perhaps, is some noise and artifacts created when I converted the file to jpeg.  I had to downscale the pixels by a third and save using the low quality of 2 in order to get a smaller file.  I have not been able to upload bigger and better resolution files - somehow it hangs up when I try to post even though at ~2MB, they are significantly smaller than the maximum allowed size.   If I look at the original scan the gradations are much smoother than the file I attached. 

[nirpat89]

It's not entirely clear to me what you are doing, but let me speculate: you have tried to calibrate your scanner using an it8.7 target and Vuescan, then have scanned the target again [and applied the calibration] to produce an image.

 
Yes.  That's correct.

There are quite a few things that can go astray while doing this. The calibration text file must be present and it's location properly identified during the calibration scan, the file generated by the calibration (scanner.icc) has to be saved in a location where it will later be accessed by Vuescan during scans requiring calibration adjustment. Those later scans must have the input->media field set to the proper value for the calibration to be applied.

That is what I am doing as well while doing the actual scans.

In addition, when scanning to produce an image you have to chose the correct white balance (regardless of whether the calibration is applied or not). This looks like a white balance problem to me. I find in Vuescan that manual selecting of the white balance works best (right click on a neutral grey patch, RGB values equal and about 120).

Here I may be doing something different.  I am using "None" for Color Balance.  There are bunch of other options that I never knew enough to play around.  I use the same when doing the calibration.  May be I will try the "Manual" mode and see it works better.

Thanks.

:Niranjan.

[Mark D Segal]

Thanks, Brad.  I appreciate your input. 

I wonder if what Mark is alluding, perhaps, is some noise and artifacts created when I converted the file to jpeg.  I had to downscale the pixels by a third and save using the low quality of 2 in order to get a smaller file.  I have not been able to upload bigger and better resolution files - somehow it hangs up when I try to post even though at ~2MB, they are significantly smaller than the maximum allowed size.   If I look at the original scan the gradations are much smoother than the file I attached.

Actually I was wondering about that too, but I don't think file resolution is the issue. What I was seeing isn't so much a problem with gradations, but of tonal neutrality between the grayscale patches. It's considerably worse with the magenta-biased image than with the green-biased one. Jay may be correct that there is a white balance issue, but that works across the board, whereas the effect bothering me is more localized. So there could be more than one thing happening.

[degrub]

How old is the lamp in the scanner ?
It may have color shifted.

[nirpat89]

How old is the lamp in the scanner ?
It may have color shifted.

Pretty old (still original.)   It may as well be the case. 

[degrub]

The other thing that can get you with older CCR tubes is adequate warm up time. Usually 20 minutes is enough. That could explain the shift between the two calibration runs.

If it is a replaceable lamp, you might be able to find a very high CRI equivalent replacement.
If i remember correctly, there is a calibration strip under the starting area of the glass that is used for internal calibration. i don't know of a replacement though if it has yellowed.