Unexpected difference between Proofing and Color Conversion

[eronn]

I want to print an image at an online service, but first I want to use a printer color profile to make sure that certain details will become visible on the print. However, there seems to be inconsistency between the the print proofing tool and the color space conversion in how the result looks on my screen. I believed that print proofing and color profile conversion should have the same effect on the sceen view. Have I missunderstood something here?

Below a detailed description of what I did:

• The document's color profile is initially set to sRGB.
• Set up proofing mode:
  Choose Options>Devices
  - Check "Activate Color Management"
  - Screen: <a scren profile measured with my Spyder tool>
  - Proof: <the printer profile>
  - Rendering Intent: Perceptual
  Click "OK"
• Activate View>Proof mode (Ctrl+2)
• Optimize the contrast for printing using Curves.

Now, I want to convert the image to the printer color space, with the profile embedded in the file. Isn't that a good strategy, before uploading to the printing service?

• Turn off proofing mode (Ctrl+2)
• Do color space conversion:
  Choose Tool>Color>Convert With Color Profile
  - Dest.Profile: <the printer profile>
  - Rendering: Perceptual
  - Destination: Document
  Click "OK"

Black areas (0,0,0) after conversion are darker on the screen than when I only used proofing. I thought they should be exactly the same. I have also tried to use Relative Colorimetric as Rendering Intent for both proofing and the profile conversion. In that case, blacks are about equal in the proof view and in the converted image. However, intensities slightly above black like (8%,8%,8%), become brighter after conversion than when proofing.

Here is the printer profile I used:
https://static.whitewall.com/ICC/WhiteW ... _Glass.icc

Erlend

[eronn]

I have investigated this a bit more.

It turns out that a single layer document/picture (rgb 16 bit) displays as expected when I choose Relative Colorimetric rendering. In this case proofing and color profile conversion result in the equal screen views. But when I choose Perceptual rendering for the same document proofing and profile conversion do still produce different views.

When I change to a multi-layered more complex document (including layers for Histogram Corrections, Curves, Saturation adjustments, Blur filters, blended layers, layer masks), the problem exists with both perceptual and the relative colorimetric rendering, as described in the previous post.

I am still confused:

• When an image is converted to the printer color spaceit can be printed "as it is", Then I wouldn't expect the proofing mode to change the view any more.
• I didn't expect proofing and color space conversion to behave differently for single layered and multilayered documents.

I suspect there is an error in the software, but I am not not experienced enough to be more than 85 % sure.

Erlend

[bkh]

Now, I want to convert the image to the printer color space, with the profile embedded in the file. Isn't that a good strategy, before uploading to the printing service?

Don't, unless the printing service specifically asks you to do so. Most printing services assume the sRGB colour space, even if the image has a different profile, so be sure to double check the service's documentation. Most proofing profiles aren't the actual profiles used for printing (this is evident in your case, since the profile is an RGB profile which is unsuitable for printing).

It turns out that a single layer document/picture (rgb 16 bit) displays as expected when I choose Relative Colorimetric rendering. In this case proofing and color profile conversion result in the equal screen views. But when I choose Perceptual rendering for the same document proofing and profile conversion do still produce different views.

This is normal. Proofing converts to the printing profile using whatever rendering you choose, and then to the monitor colour space using relative colorimetric (to show the actual printer colours). There are test profiles on the ICC web page which can be used to verify this – last time I checked, PL behaved that way. Try taking a screenshot of the proof using "Perceptual" and one of the document, converted to the proofing profile using "Perceptual", but viewed in PL with "Relative Colorimetric" as the monitor setting. These should be the same.

When I change to a multi-layered more complex document (including layers for Histogram Corrections, Curves, Saturation adjustments, Blur filters, blended layers, layer masks), the problem exists with both perceptual and the relative colorimetric rendering, as described in the previous post.

If you convert your document to another colour space, you are effectively changing the working colour space of all the adjustment layers (except for child layers and isolated groups, which I think may have their own working space, but that's a different story). It's a basic rule that you choose your working colour space first, then make adjustments. The bigger the difference in the colour spaces, the bigger the effects on adjustment layers.

Therefore, as far as I can see, PL works as it's supposed to.

Cheers

Burkhard.

[eronn]

Thank you Burkhard for the reply!

Most printing services assume [*]the sRGB colour space, even if the image has a different profile, so be sure to double check the service's documentation. Most proofing profiles aren't the actual profiles used for printing (this is evident in your case, since the profile is an RGB profile which is unsuitable for printing).

As far as I understand the print service will use embedded color profiles. From the Whitewall FAQ pages: "Your files may be uploaded in RGB or in CMYK. CMYK files will be converted into RGB in the production process. If a colour profile is not attached to the file, we will assume that your file has an sRGB profile".

I want to explicitely control the color conversion and to exploit the full gamut of the print process. The ICC standard ensures that this can be achieved just as precisely with a device specific RGB space as with a CMYK space. In general, sRGB does not cover all colors that can be reproduced by good printers (especially saturated greens and some yellows). I am not concerned about the remaining data conversions that is technically needed for the printing process. (For instance, a CMYK printer can often reproduce a dark color with only color inks, or by including some black ink. I just want the printing service to use their default settings for these details).

Proofing converts to the printing profile using whatever rendering you choose, and then to the monitor colour space using relative colorimetric (to show the actual printer colours). There are test profiles on the ICC web page which can be used to verify this – last time I checked, PL behaved that way. Try taking a screenshot of the proof using "Perceptual" and one of the document, converted to the proofing profile using "Perceptual", but viewed in PL with "Relative Colorimetric" as the monitor setting. These should be the same.

I tested three different versions, and they all look different, in contrast to the expectation:

• document proofed with Perceptual
• document converted with Perceptual, proofed with colorimetric rendering
• document converted with Perceptual, no proofing, but Color Management rendering set to colorimetric

I tried to make screen shots (using the keyboard PrintScreen button), and they also looked different. However, the colors are not preserved in the screen shot copy process. Pasting a screen shot into a new PL document changes the colors when color management is active (tested with and without proof mode after pasting). If you want, I can try to illustrate with some photos of my screen instead.

(BTW: The PL documentation should maybe explain which kind of rendering is used for each of the conversions in the Color Management dialog, and in particular, which of the conversions that the rendering specified at the bottom of the dialog applies to.)

If you convert your document to another colour space, you are effectively changing the working colour space of all the adjustment layers [...]. It's a basic rule that you choose your working colour space first, then make adjustments.

I see! I didn't know that this function simply converted each layer individually. Conversion layer is obviously not a desirable approach in most cases where color adjustment layers etc. are included. There should be a warning about this in the documentation, since novice users like me might not be aware that there is a difference between converting before or after flattening the document.

Therefore, as far as I can see, PL works as it's supposed to.

Not quite (on my computer) when it comes to proofing versus conversion of single layered documents in the perceptual mode. In this case I expect identical results. It also makes me sceptic that that an image that has allready been converted to the printing profile changes colors when I proof it with the same printing profile.

Cheers,
Erlend

[bkh]

In general, sRGB does not cover all colors that can be reproduced by good printers (especially saturated greens and some yellows). I am not concerned about the remaining data conversions that is technically needed for the printing process.

In that case, any colour space at least as large as the printing colour space will do (including the proof profile). But even the RGB proofing profile supplied by your photo service will very likely contain unprintable colours, so you'll have to proof your photos anyway.

Proofing converts to the printing profile using whatever rendering you choose, and then to the monitor colour space using relative colorimetric (to show the actual printer colours). There are test profiles on the ICC web page which can be used to verify this – last time I checked, PL behaved that way. Try taking a screenshot of the proof using "Perceptual" and one of the document, converted to the proofing profile using "Perceptual", but viewed in PL with "Relative Colorimetric" as the monitor setting. These should be the same.

I tested three different versions, and they all look different, in contrast to the expectation:

• document proofed with Perceptual
• document converted with Perceptual, proofed with colorimetric rendering
• document converted with Perceptual, no proofing, but Color Management rendering set to colorimetric

Works here (OS X), using the ICC test profile (Probe1_ICCv2, available from http://www.color.org/probeprofile.xalter). I suggest you try the test profile yourself – they are made in such a way that different rendering intents create totally different output. Also tried your proofing profile, no difference there, either.

However, the colors are not preserved in the screen shot copy process.

Probably, your screenshot software doesn't attach the monitor profile to the screenshots. Anyway, the three screenshots should look the same (you could turn off colour management in PL to see just the monitor colours).

(BTW: The PL documentation should maybe explain which kind of rendering is used for each of the conversions in the Color Management dialog, and in particular, which of the conversions that the rendering specified at the bottom of the dialog applies to.)

Better still, I'd wish that PL allowed to specify one rendering intent for each device …

If you convert your document to another colour space, you are effectively changing the working colour space of all the adjustment layers [...]. It's a basic rule that you choose your working colour space first, then make adjustments.

I see! I didn't know that this function simply converted each layer individually. Conversion layer is obviously not a desirable approach in most cases where color adjustment layers etc. are included. There should be a warning about this in the documentation, since novice users like me might not be aware that there is a difference between converting before or after flattening the document.

Not quite (on my computer) when it comes to proofing versus conversion of single layered documents in the perceptual mode.

As I wrote before, the difference is that when converting your document, screen output depends on the rendering intent used by the display profile (i.e., the intent set under Preferences -> CM -> Devices) while during proofing, display conversion is always relative colorimetric.

It also makes me sceptic that that an image that has allready been converted to the printing profile changes colors when I proof it with the same printing profile.

This shouldn't happen if you use "relative colorimetric" for the display and don't use out-of-gamut colours. (As I wrote above, the RGB space of the proof profile is likely larger than the printer's real device space.) Perceptual rendering on the screen can be different from proof mode, that's perfectly normal.

Cheers

Burkhard.