Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Shifting greys with different rendering intents  (Read 5961 times)
Rhossydd
Sr. Member
****
Offline Offline

Posts: 1889


WWW
« on: May 30, 2007, 12:21:39 PM »
ReplyReply

Can someone help me understand what’s started to go on here ?
I’ve started to see differences on grey scales when swapping between relative colormetric and perceptual rendering intents with recent profiles made with Profilemaker Pro 5 and my Eye-One Pro.

As I understand it, both should render neutral greys the same, as some profiles do.

However if I print the same grey scale with both intents I’m seeing a colour shift between them, varying between 2.1 and 4 DeltaE in dark tones(L 5&20) going up to dE of around 6 for lighter tones (L 32-91).

Is this a sign of a dying Eye-One ? profiles made a year or two ago show no such behaviour. I’ve run Eye-One diagnostics, but it returns a pass for it.
Anyone seen anything like this before ?

Thanks in advance.

Paul
Logged
Rhossydd
Sr. Member
****
Offline Offline

Posts: 1889


WWW
« Reply #1 on: June 02, 2007, 01:39:19 AM »
ReplyReply

No one else seen anything like this then ?

GMB say colour shifts in greys are to be expected when using perceptual RI, that seems wrong to me.
Logged
Ernst Dinkla
Sr. Member
****
Offline Offline

Posts: 2806


« Reply #2 on: June 02, 2007, 03:13:53 AM »
ReplyReply

Quote
No one else seen anything like this then ?

GMB say colour shifts in greys are to be expected when using perceptual RI, that seems wrong to me.
[a href=\"index.php?act=findpost&pid=120748\"][{POST_SNAPBACK}][/a]


First of all what kind of printer ?  A CMYK, CcMmYK or one that can build the grey tone range with grey inks ?
Second how neutral is your profile in the rendering you intended to use it for, deviations between prints with the same rendering intent ?
Third, did you try the profile in more than one application, color engine and by that in more rendering and BPC interpretations ?

Ernst Dinkla

www.pigment-print.com
Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2101


« Reply #3 on: June 02, 2007, 11:11:52 AM »
ReplyReply

Hmm, are you choosing paper gray or neutral gray for your perceptual options in ProfileMaker?

Paper gray will shift the gray axis throughout the entire tonal scale from shadows to highlights (media-relative colorimetry).

Neutral gray will shift the gray axis for the highlight region (media-relative colorimetry) but not shift the gray axis in lower areas of the tonal scale (i.e., shadows); this corresponds to absolute colorimetry with black point compensation in the shadows.

Eric
Logged

Rhossydd
Sr. Member
****
Offline Offline

Posts: 1889


WWW
« Reply #4 on: June 05, 2007, 01:05:23 AM »
ReplyReply

To add the extra detail asked for;

I'm seeing this behaviour on RGB profiles created for normal desktop inkjet printers. I've seen the same issue on profiles made for a range of different Epson and Canon printers.
I've tried both paper and neutral grey options and both display a similar behaviour. although paper grey axis profiles are sometimes marginally better.
The colour shift is seen when printing from any colour managed application I've tried, PS CS2, CS3, LR, Qimage and can be seen when soft proofed as well as when printed. So I assume the CME is not a factor in this issue.

My main question is; Should one see a difference in grey scales when printed with the different intents at all ?
My understanding is that rendering intents should remap just colour within the gamut and leave the grey axis alone.
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 8628



WWW
« Reply #5 on: June 05, 2007, 08:27:45 AM »
ReplyReply

A percpetual intent may indeed alter grays, also, out of gamut isn't exclusive to colors but tone (grays) too. Lastly, there's no standards in how a profile manufacturer can code how they want to build a perceptual mapping. The GMB profiles may show different behavior based on the user settings in how this table is built and the math used is up to the profile builder, they can render the mapping as they think you'll prefer.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Rhossydd
Sr. Member
****
Offline Offline

Posts: 1889


WWW
« Reply #6 on: June 05, 2007, 02:21:22 PM »
ReplyReply

Quote
A percpetual intent may indeed alter grays, also, out of gamut isn't exclusive to colors but tone (grays) too.

I can understand why a grey scale might be tonally compressed or extended to fit the grey scale gamut of the materials, but I fail to understand why a grey would have a colour cast added to it.

Have you seen behaviour like this with your own profiles ?
If so, how would you justify it ?
Logged
rdonson
Sr. Member
****
Offline Offline

Posts: 1420


WWW
« Reply #7 on: June 05, 2007, 06:27:36 PM »
ReplyReply

Quote
I can understand why a grey scale might be tonally compressed or extended to fit the grey scale gamut of the materials, but I fail to understand why a grey would have a colour cast added to it.

Have you seen behaviour like this with your own profiles ?
If so, how would you justify it ?
[a href=\"index.php?act=findpost&pid=121292\"][{POST_SNAPBACK}][/a]

According to your original post you're only seeing these shifts in your prints, not in softproofing.  Is this just the difference between desktop printers of the same manufacture?  Which printers?  Are these printers calibrated?  Linearized?  A color cast suggests a less than perfect profile or a printer drifting.
Logged

[span style='font-size:14pt;line-height:100%'][span style='font-family:Arial'][span style='font-family:Geneva'][span style='font-size:8pt;line-height:100%']Regards,
Ron[/span][/span][/span][/span]
Rhossydd
Sr. Member
****
Offline Offline

Posts: 1889


WWW
« Reply #8 on: June 06, 2007, 01:37:08 AM »
ReplyReply

Quote
According to your original post you're only seeing these shifts in your prints, not in softproofing.
No, I didn't mention soft proofing in my first post at all.
Quote
A color cast suggests a less than perfect profile.
Indeed, but where is the error originating from ?

Are GMB right to suggest that greys should change colour when rendering intents are changed ?
It doesn't sound correct to me.
Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2101


« Reply #9 on: June 06, 2007, 08:49:16 AM »
ReplyReply

Quote
Are GMB right to suggest that greys should change colour when rendering intents are changed ?
It doesn't sound correct to me.
[a href=\"index.php?act=findpost&pid=121359\"][{POST_SNAPBACK}][/a]

It's a complicated issue, but yes, it is reasonable that a profile might shift the gray axis among the different intents.

Imagine that the darkest patch that your printer can print on a given paper isn't neutral (it rarely is). Then the question is, how should this patch be handled by a profile? Is this considered neutral black, or is it considered a non-neutral color that happens to be darker than the darkest neutral gray? In other words, the point I'm making is that there is freedom in interpretation of exactly "what is the gray axis". There's not a clear-cut answer here. If you want the black (LAB (0,0,0)) to map to the darkest neutral gray in the print, then indeed you may get very neutral grays but then have a disappointing d-max. Conversely, if a profile maps LAB (0,0,0) to the darkest tone the printer can produce, neutral or not, then you'll get excellent d-max but the shadow grays may get shifted to be non-neutral in order to make this happen. As I said, it's complicated.

The short answer to your question is that, yes, it is entirely conceivable that the definition of the gray axis may change between the various intents.

The practical advice is to make test prints and determine which best suits your image ...
Logged

Rhossydd
Sr. Member
****
Offline Offline

Posts: 1889


WWW
« Reply #10 on: June 07, 2007, 01:37:34 AM »
ReplyReply

Quote
Imagine that the darkest patch that your printer can print on a given paper isn't neutral (it rarely is). Then the question is, how should this patch be handled by a profile?
The short answer to your question is that, yes, it is entirely conceivable that the definition of the gray axis may change between the various intents.

I can see the point you're making, although I remain to be convinced it's the correct behaviour and contradicts some of the statements made by other authorities who suggest that there shouldn't be any change in grey scale on rendering intents.

However, my problem isn't an issue of dark shadows shifting, but the lighter tones. From the DeltaE figures I quoted in my first message you'll see that as the luminanace value decreases, the difference decreases as well.

Quote
The practical advice is to make test prints and determine which best suits your image ...

I'm not interested in 'working round' the problem, but curing it.
So far no one else has said they're seeing this issue with profiles they've built and the OEM profiles I have here all show no shift between perceptual and RC RI, so it seems to be a problem with profiles I've built here.
I'm just trying to identify the source of the problem, which seems at the moment to be most likely a failing Eye-One.
Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2101


« Reply #11 on: June 07, 2007, 05:46:47 AM »
ReplyReply

Quote
I can see the point you're making, although I remain to be convinced it's the correct behaviour and contradicts some of the statements made by other authorities who suggest that there shouldn't be any change in grey scale on rendering intents.

Depends on what you mean by the "correct" behavior. Having the neutral axis be the same between different rendering intents is not specified in any way by the ICC specification.

I have just opened a 256-patch 8-bit gray test image and tried converting it to different printer space using profiles built with various tools. This includes ProfileMaker 5 profiles (built using the Paper Gray option), MonacoPROFILER profiles, and Epson-supplied profiles for the 3800 (not sure what tool they use, but I don't think it's either of these). With all sets of profiles, regardless of how the profile is built, the gray axis is rendered differently among the different intents. This is easy to see because the actual device RGB values change when I change the Rendering Intent. In some cases it's a small change and may not be perceivable. In some cases the change is bigger. MonacoPROFILER profiles, for instance, add a fairly big tonal boost in the highlights for the perceptual tables, compared to the relative colorimetric table.

Quote
However, my problem isn't an issue of dark shadows shifting, but the lighter tones. From the DeltaE figures I quoted in my first message you'll see that as the luminanace value decreases, the difference decreases as well.
I'm not interested in 'working round' the problem, but curing it.
So far no one else has said they're seeing this issue with profiles they've built and the OEM profiles I have here all show no shift between perceptual and RC RI, so it seems to be a problem with profiles I've built here.
I'm just trying to identify the source of the problem, which seems at the moment to be most likely a failing Eye-One.
[a href=\"index.php?act=findpost&pid=121537\"][{POST_SNAPBACK}][/a]

It is true that the shift you're seeing is a lot bigger than what's expected (sorry, glossed over your DeltaE values the first time). Hmm. I assume you're using the same target files and PMP software options as previously? Are you sure your calibration tile is reasonably clean? Do you have a clean color checker that you can measure and compare to the reference values?
Logged

Chris_T
Sr. Member
****
Offline Offline

Posts: 541


« Reply #12 on: June 07, 2007, 06:50:59 AM »
ReplyReply

Quote
Lastly, there's no standards in how a profile manufacturer can code how they want to build a perceptual mapping.
[a href=\"index.php?act=findpost&pid=121226\"][{POST_SNAPBACK}][/a]

So true. While many books and tutorials provide detailed descriptions of how each intent is *supposed" to behave, few if any profile manufacturers or vendors will tell you how they *actually* code the profiles' intents. You end up getting profiles that will make some difference, but may not be what you expect.
Logged
Chris_T
Sr. Member
****
Offline Offline

Posts: 541


« Reply #13 on: June 07, 2007, 06:56:11 AM »
ReplyReply

Quote
So true. While many books and tutorials provide detailed descriptions of how each intent is *supposed" to behave, few if any profile manufacturers or vendors will tell you how they *actually* code the profiles' intents. You end up getting profiles that will make some difference, but may not be what you expect.
[a href=\"index.php?act=findpost&pid=121575\"][{POST_SNAPBACK}][/a]

In addition, how gamut warning in PS CS Soft Proofing is implemented may also be up to the discretion of the coders. Often, I find while Soft Proofing with a media profile with gamut warning turned on, nothing would be flagged but the prints will come out with tonal/color differences.
Logged
Chris_T
Sr. Member
****
Offline Offline

Posts: 541


« Reply #14 on: June 07, 2007, 07:21:31 AM »
ReplyReply

Quote
I can see the point you're making, although I remain to be convinced it's the correct behaviour and contradicts some of the statements made by other authorities who suggest that there shouldn't be any change in grey scale on rendering intents.
[{POST_SNAPBACK}][/a]

I too was led by many "authorities" to believe that if I invest in calibration tools and custom profiles, I can get prints that will *precisely* match monitor images. After many frustrating hours questioning my equipment and/or technique inadequacy, I finally come to grips that my expectation was simply unrealistic.

More recently, some are coming out with a more realistic approach:

[a href=\"http://www.johnpaulcaponigro.com/lib/downloads/#]http://www.johnpaulcaponigro.com/lib/downloads/#[/url]

Look for the Bracket Proofing pdf under Techniques - Proofing.

Add Eric's suggestion below, and we are pretty much back to printing in a traditional darkroom.

"The practical advice is to make test prints and determine which best suits your image ... "

May those "authorities" rot in hell.
Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2101


« Reply #15 on: June 07, 2007, 04:51:46 PM »
ReplyReply

I agree, your expectations were unrealistic. The print can't match the monitor display "exactly" for several reasons:

1. differences in absolute dynamic range, which affect contrast perception

2. differences in color gamut; what happens when a color that can be printed cannot be displayed on the monitor? no way to preview it!

3. differences in adaptation (e.g., white point of display doesn't match white point of paper, and even though supposedly the user will adapt, it's imperfect)

4. possible metameric effects

(this is assuming perfect profiles, which of course don't exist either)

The main idea is that you can get very good consistency and predictability, even if the match is not quite exact. That is, with some experience and a well-controlled system, one can learn what to expect from the output.
Logged

Schewe
Sr. Member
****
Online Online

Posts: 5425


WWW
« Reply #16 on: June 07, 2007, 07:27:14 PM »
ReplyReply

Quote
I too was led by many "authorities" to believe that if I invest in calibration tools and custom profiles, I can get prints that will *precisely* match monitor images....

...May those "authorities" rot in hell.
[a href=\"index.php?act=findpost&pid=121580\"][{POST_SNAPBACK}][/a]

Well, I've never heard any of the mainline color management experts I know (and I know them all) ever say color managment will allows you match a print and display *precisely*...I've heard experts say (and I've said it too) that having accurate profiles and knowing how to use Photoshop's softproofing allows you to get about a 90% accurate prediction on screen of what the print will look like...and I'll hold that line...but it requires an accurate profile for the display, the printer/paper and knowing how to use soft proofing...THEN you make a print and see if ya gotta tweak....that last 10%.

With today's displays, wider gamut printers and profile making capability, I think 90% is pretty good...it sure saves a lot of ink and paper and allows far better control over the final image.
Logged
Chris_T
Sr. Member
****
Offline Offline

Posts: 541


« Reply #17 on: June 12, 2007, 07:15:07 AM »
ReplyReply

I do understand the differences, and perhaps my use of "exactly" is unrealistic.

With proper color management, you can get close. But nothing beats reviewing test prints side by side and pick the one you like the best.

In a traditional darkroom, one way is to print a sheet with different exposures across it. That simplies the reviewing process somewhat. In a digital darkroom, a similar action that can print one sheet with different tonal/color variations can be very useful. I recall that there was such a product a while back, but it does not seem to be around anymore.

Quote
I agree, your expectations were unrealistic. The print can't match the monitor display "exactly" for several reasons:

1. differences in absolute dynamic range, which affect contrast perception

2. differences in color gamut; what happens when a color that can be printed cannot be displayed on the monitor? no way to preview it!

3. differences in adaptation (e.g., white point of display doesn't match white point of paper, and even though supposedly the user will adapt, it's imperfect)

4. possible metameric effects

(this is assuming perfect profiles, which of course don't exist either)

The main idea is that you can get very good consistency and predictability, even if the match is not quite exact. That is, with some experience and a well-controlled system, one can learn what to expect from the output.
[a href=\"index.php?act=findpost&pid=121697\"][{POST_SNAPBACK}][/a]
Logged
Chris_T
Sr. Member
****
Offline Offline

Posts: 541


« Reply #18 on: June 12, 2007, 07:28:23 AM »
ReplyReply

For those who strive for the best results (in just about anything), getting that last 5 to 10% "right" typically takes 90 to 95% of the effort. But aside from the creators, few will or can notice the difference.

The pros who need to meet a deadline or make a living are the ones that love the quick 90% solutions. And there's nothing wrong with it.

Quote
Well, I've never heard any of the mainline color management experts I know (and I know them all) ever say color managment will allows you match a print and display *precisely*...I've heard experts say (and I've said it too) that having accurate profiles and knowing how to use Photoshop's softproofing allows you to get about a 90% accurate prediction on screen of what the print will look like...and I'll hold that line...but it requires an accurate profile for the display, the printer/paper and knowing how to use soft proofing...THEN you make a print and see if ya gotta tweak....that last 10%.

With today's displays, wider gamut printers and profile making capability, I think 90% is pretty good...it sure saves a lot of ink and paper and allows far better control over the final image.
[a href=\"index.php?act=findpost&pid=121711\"][{POST_SNAPBACK}][/a]
Logged
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad