Black rendering with different intents

[juicy]

Hi!

Could someone elaborate on why some printer profiles have very weak black when using relative colorimetric rendering intent w blackpoint compensation compared to the same profile/paper combination with either perceptual or relative without bp comp? Is this because relative colorimetric w bpc maps the source black to the darkest neutral color the media can achieve and perceptual maps source black to the darkest color (max density, neutral or not) the media can achieve or is it perhaps because of a bad profile?

I've seen this especially with some canned profiles for matte papers and the differences are sometimes huge but also to much lesser extent with some custom profiles for E7800/premiumgl. The effect is seen similarly in print and softproof and the numbers going to the printer are very different (extreme case: source 0,0,0 -> relcol bpc 20,20,21 compared to perceptual 0,0,0).

Thanks in advance,
J

[Mark D Segal]

Your question is a mixture of three sets of variables: Rendering Intent, Black Point Compensation (BPC) and choice of paper type. It's hard to tell whether your problem is due to your profiles or your management of the soft-proof and print options unless you disaggregate your description of what happens under each set of conditions which you can select. For soft-proofing, I'll throw-in a fourth important one: "Simulate Paper Color" (SPC). Use your soft-proof function to carefully observe what happens with each paper, rendering intent and choice of BPC and SPC, then provide here a careful description of what you see under each set of choices, and then I can tell you whether your observations are consistent with mine under the same conditions. That would tell you whether the issue is settings or profiles.

[juicy]

Your question is a mixture of three sets of variables: Rendering Intent, Black Point Compensation (BPC) and choice of paper type. It's hard to tell whether your problem is due to your profiles or your management of the soft-proof and print options unless you disaggregate your description of what happens under each set of conditions which you can select. For soft-proofing, I'll throw-in a fourth important one: "Simulate Paper Color" (SPC). Use your soft-proof function to carefully observe what happens with each paper, rendering intent and choice of BPC and SPC, then provide here a careful description of what you see under each set of choices, and then I can tell you whether your observations are consistent with mine under the same conditions. That would tell you whether the issue is settings or profiles.
[a href=\"index.php?act=findpost&pid=205668\"][{POST_SNAPBACK}][/a]

Thanx for your comments.

This is an example of the most extreme case.
Attached is a comparison of 2 screengrabs (with some jpeg artifacts of course  ) in one image showing the softproof settings and a test image containing black and dark grey gradient. Sourcespace was sRGB for this comparison. Lefthand readings in colorsampler window is actual color (sRGB 0,0,0) and righthand reading is proof color. Canon's cheap matte photopaper and their canned profile for this paper, CS2, osx 10.4.11, screen profiled with Optix XRpro.

The problem is not that softproof is inaccurate, on the contrary, even with a cheap iP5300 printer I get quite consistent screen-to-print-match. The problem is (especially with this paper-profile-combination) that peceptual rendering induces some unwanted lightening of the image and some hue shifts. On the other hand relative rendering w bpc does not allow maximum black on this paper (see the difference in the attached image). Now, I can live with this situation because I don't print anything meaningful with this printer-paper-combination anyway but I just noticed similar behavior (although to an almost negligible extent) with a custom profile made for Eps7800.

Question is more about the way black or maximum density is normally achieved in a given profile.

Cheers,
J

[Mark D Segal]

Perceptual rendering is meant to cause more hue shift than is RelCol, because of the different way it handles OOG colours. (The print module roll-over dialogs in CS3 10.0.1 explain quite nicely how each of these rendering intents handles colours.)

In your illustration, you have BPC turned off in one case and turned on in the other. Where it is turned on you get a more accurate view of the paper black you will see in the print, based on what the profile is telling Photoshop about the quality of black in the paper for which that profile is intended - and in this case being cheap matte paper, the quality of the black will not be as deep as it is in the file or on your display (without BPC) because the DMax of the paper is simply not that spectacular.

I don't see anything abnormal about any of this.

[juicy]

Perceptual rendering is meant to cause more hue shift than is RelCol, because of the different way it handles OOG colours. (The print module roll-over dialogs in CS3 10.0.1 explain quite nicely how each of these rendering intents handles colours.)

In your illustration, you have BPC turned off in one case and turned on in the other. Where it is turned on you get a more accurate view of the paper black you will see in the print, based on what the profile is telling Photoshop about the quality of black in the paper for which that profile is intended - and in this case being cheap matte paper, the quality of the black will not be as deep as it is in the file or on your display (without BPC) because the DMax of the paper is simply not that spectacular.

I don't see anything abnormal about any of this.
[a href=\"index.php?act=findpost&pid=205707\"][{POST_SNAPBACK}][/a]

Thanx again.

Black point compensation is turned off in perceptual rendering because there is no black point compensation option in perceptual rendering tables. Whether you tick the bpc box or not that does not have any kind of effect with perceptual rendering (bpc does not exist with perceptual). With relative however, bpc is the key to preserve shadow detail.

There is certainly nothing abnormal in getting hue shifts with perceptual rendering and that was not my point at all. I just mentioned it because that is the reason why I would prefer relative colorimetric rendering (on this paper) if only it would give me the same dmax as perceptual rendering on the very same paper.

This is not a problem of softproofing.   The question is why perceptual rendering produces astronomically better blacks on the same paper-printer-combination?

Cheers,
J

[Mark D Segal]

Yes, you are largely correct that BPC doesn't affect a Perceptual Rendering Intent (it does just a very wee bit on the Atkinson Printer Test Target for the Epson Pro38 EMP profile, but not for the Ilford GFS profile), and it is correct that RelCol preserves better tonal separation using BPC. However, testing the choice between Perceptual (ignoring BPC) and RelCol with BPC, the quality of black is unaffected  between these options using the several profiles with which I've tested it (Pro38 EMP, Pro38 PLPP and Ilford GFS). Hence, based on my observation, the choice of Perceptual should not be producing "astronoimically better blacks"; this leads one to suspect a profile issue specific to the profile you are using for that paper.

[juicy]

*
*
 However, testing the choice between Perceptual (ignoring BPC) and RelCol with BPC, the quality of black is unaffected  between these options using the several profiles with which I've tested it (Pro38 EMP, Pro38 PLPP and Ilford GFS). Hence, based on my observation, the choice of Perceptual should not be producing "astronoimically better blacks"; this leads one to suspect a profile issue specific to the profile you are using for that paper.
[a href=\"index.php?act=findpost&pid=205735\"][{POST_SNAPBACK}][/a]

With most profiles I have used for various output there has not been this "diluted black" problem.

For some reason however the same problem I see with Canon's canned profiles I see to some extent for example with Ilford smooth fa paper with their profile. The numbers to the printer are 0,0,0 with perceptual but 5,19,0 with relative w bpc and the difference in dmax is obvious in the print (although nowhere near as bad as with the aforementioned Canon paper).

This led me to the original question whether relative rendering is based on achieving the best possible *neutral* dmax? And perhaps perceptual is whatever the profiler thinks looks good. In the case of this particular printer there might be no way to get a decent neutral black but with perceptual rendering it's possible to mix whatever ink densities needed for maximum dmax with obvious tint in the black as a side effect? Does this make any sense?

Cheers,
J

(Anyway, sorry for possibly cryptic post, this is not my native language.)

[Mark D Segal]

RelCol compares the white of the source with the white of the destination colour space and adjusts all colours accordingly. It brings OOG colours to the closest reproducible colour in the destination space. It preserves more of the original image colours than would perceptual. Perceptual preserves visual relationships between colours. I can't say how the math does this, but one must presume the profiling packages are based on considerable testing and experience with human visual perception. By the time one gets to BLACK, however, one wouldn't think the choice of RelCol with BPC should differ from Perceptual, so again, I think you have a profile problem - you won't get L=0 in a print, but the non-neutrality of the printed black relative to the neutrality of the file numbers really suggests that either your profile is no good or the printer is not working properly.

[juicy]

*
*
By the time one gets to BLACK, however, one wouldn't think the choice of RelCol with BPC should differ from Perceptual, so again, I think you have a profile problem - you won't get L=0 in a print, but the non-neutrality of the printed black relative to the neutrality of the file numbers really suggests that either your profile is no good or the printer is not working properly.
[a href=\"index.php?act=findpost&pid=205772\"][{POST_SNAPBACK}][/a]

I'm convinced it's a profile issue because softproof shows exactly what happens in print.

I guess I have to have a custom profile made for that matte paper just to see what happens.

Cheers,
J