Printing "rendering intents" and shadow separation

[Marco B]

Hi all,

Remark:
Before you start commenting: please take notice of the fact that I have spend MUCH time investigating and learning colormanagement, and am pretty much sure I understand the basics well now and know how to apply it in colormanaged software like Photoshop. My monitor is properly calibrated using a Spider 2.

The thread topic:
Searching for some other stuff, I came past this webpage:

http://www.pbase.com/petolino/image/55994572

It contains an image that consist of two area's: a completely white (RGB 0,0,0) and black (RGB 255,255,255) area. Both contain numbers, the value of which actually represent the RGB value of the grey value that constitutes the fonts color.

That may sound a bit difficult, but the purpose of this small test chart is to test highlight and shadow performance of your printing system. See the text below the image on the webpage for details on how to use it.

Now by opening it in Photoshop, asigning the AdobeRGB colorspace, and setting the "Proof Setup" in Photoshop for different paper types using manufacturer supplied ICC profiles, and using both the rendering intents "Perceptual" and "Relative Colorimetric", I noticed there is actually a pretty huge and slightly unexpected difference in the shadow rendering performance between the two rendering intents, with in most cases (but not all!) "Perceptual" performing better. Please note that "Paper Color" simulation was on in both cases, as can be witnessed by the greyish matte paper proofings.

And YES, I did switch ON "Black point compensation" too for "Relative Colorimetric"! (is by default on for Perceptual, because it's and integral part of that intent, although Photoshop doesn't show it. An excellent reference to the function of black point compensation is here: Gamutvision - Black Point Compensation)

To check this observation, I also printed on Harman FB Matt Warmtone using both rendering intents and Harman's ICC profile for this paper and letting Photoshop handle the colormanagement. The prints from my EPSON R2400 were consistent with the screen observations, just a little bit gain, meaning 1 or 2 scales extra visible in the print compared to the screen.

The difference is especially apparent in matte papers. For example, see the image below of Hahnemuhle PhotoRag, it doesn't separate any of the values below about RGB 20,20,20 with Relative Colorimetric, while it goes down to RGB 3,3,3 with Perceptual. Please note that you will need to strain hard to see these darkest values! I converted these screenshots from my monitor- to the sRGB-profile, but it will be better if you download the test chart yourself and set proofing using the profiles of the papers in Photoshop.

Using the "Perceptual" intent seems to be giving far better shadow separation than "Relative Colorimetric". Now this slightly surprises me, because the general consensus on many webpages seems to be that the difference is small, and that "You just need to try it and see which is best...", with many people opting for Relative Colorimetric too...

And "Black Point Compensation" used with Relative Colorimetric should (and will alleviate) part of the problem that exists with Relative Colorimetric potentially clipping deepest shadows (see the article mentioned above), but not to the extent I would have expected it.

Anyway, my questions to you:

• Has anyone else noticed these differences?
• What rendering intent do you prefer, or how do you decide which one to use during printing?
• Any comments or good references to the rendering intents that might explain this difference better? I am especially intrigued that in at least one case (Harman FB Glossy paper) the performance of the Relative Colorimetric was not worse than Perceptual, meaning it's not just simply a matter of the choice of rendering intent alone and how that effects how image RGB values are translated in output device results (whether printer or screen), but has to do something with the ICC profiles as well.

PLEASE ALSO NOTE THE FOLLOWING:
I also noticed that the canned profiles for the EPSON papers that came with my R2400 actualy DO NOT do a "Black Ink" simulation, as the other profiles do. They also showed, probably as a consequence of this, NO difference between the proofing of the "Perceptual" and "Relative Colorimetric" rendering intents. It therefor seems likely that "Black Ink" simulation must be part of the ICC profile definition in order to properly see and proof these differences in rendering intent - and final print result!

Just look at the images below that represent the proofed images, captured as screenshots in Photoshop CS2:

HAHNEMUHLE PHOTO RAG:
Perceptual: Deepest black visible: 3 (= RGB 3,3,3)
Perceptual: Brightest white visible: 253
Relative Colorimetric: Deepest black visible: 20
Relative Colorimetric: Deepest black visible: 253

HARMAN FB GLOSSY:
Perceptual: Deepest black visible: 11
Perceptual: Brightest white visible: 251
Relative Colorimetric: Deepest black visible: 4
Relative Colorimetric: Deepest black visible: 249

HARMAN FB GLOSSY WARMTONE:
Perceptual: Deepest black visible: 5
Perceptual: Brightest white visible: 251
Relative Colorimetric: Deepest black visible: 10
Relative Colorimetric: Deepest black visible: 249

HARMAN FB MATTE:
Perceptual: Deepest black visible: 4
Perceptual: Brightest white visible: 251
Relative Colorimetric: Deepest black visible: 6
Relative Colorimetric: Deepest black visible: 249

HARMAN FB MATTE WARMTONE:
Perceptual: Deepest black visible: 5
Perceptual: Brightest white visible: 251
Relative Colorimetric: Deepest black visible: 9
Relative Colorimetric: Deepest black visible: 249

ILFORD GOLD FIBRE SILK:
Perceptual: Deepest black visible: 2
Perceptual: Brightest white visible: 252
Relative Colorimetric: Deepest black visible: 8
Relative Colorimetric: Deepest black visible: 252

ILFORD SMOOTH PEARL PAPER:
Perceptual: Deepest black visible: 5
Perceptual: Brightest white visible: 252
Relative Colorimetric: Deepest black visible: 16
Relative Colorimetric: Deepest black visible: 252

 

[Marco B]

Well, over thinking it all a bit more and re-reading some of the stuff on colormanagement I have collected over the years, I think I can now answer this question myself:

- Relative colorimetric is of course designed to maintain as much as possible the original RGB values color appearance within an image when mapped from one colorspace / gamut (e.g. the images, e.g. AdobeRGB), to another (e.g. a printer colorspace / gamut). Any image color outside a printing papers gamut is clipped to the nearest color inside the printer papers gamut. This maintains as much as possible maximum color vibrancy or color saturation, at the expense of a possible clipping of colors, that may not be visible in the final print. If the input image's colors only constitute a small gamut, no clipping might occur.

Now, black point compensation has probably been designed to compensate or "bug-fix" where the clipping of colors is most likely to occur or hurt: the darker, unsaturated black parts of the image. As can be seen from the images in the article mentioned in my previous post, this is accomplished by bending the curve that depicts input and output color at it's toe and have it more or less asymptotically reach the darkest black in the output printer gamut. However, since this is a non-linear curved approach to darkest black, the darkest colors only slowly reach the darkest 0,0,0 RGB value, causing little distinction or separation of the darkest values.

- Now contrary to this, Perceptual rendering intent matches or scales all of the colors in the input gamut to the output printing gamut. This maintains as much detail as possible over the whole brightness region from deepest blacks to brightest white and within colors, however at the possible expense of a loss of some color saturation or vibrancy, some small shifts in color appearance and maybe a bit of midtone compression. The darkest values probably reach RGB 0,0,0 in a more steep curve than "Black point compensation" does for Relative colorimetric, allowing a better deep black separation.

Now my conclusions of all this:

- In case you want to maintain maximum deep shadow separation, you are probably better of using Perceptual rendering intent if the output colorspace or printing paper gamut is small, as in the case of Matte papers.

- If the output colorspace or printing papers gamut is relatively big, as with high Glossy papers, Relative colorimetric probably does a reasonable job as well with shadow separation, although Perceptual might still gain a little.

This also, at least seems to be, consistent with all of the observations I made on the proofed images, including a not very big difference in deep shadow separation between the rendering intents on Glossy papers...

Does this make any sense, or am I rambling here?

Marco

 

[scovell001]

Hi Marco,

Had a quick flick through your text. And downloaded this image and spent some time with it. Its a really bloody useful image and I now use it as part of my profile testing all the time.

A couple of quick takes for you.

1) don't assign or convert the image to Adobe RGB before you start soft proofing, as you are adjusting the tonal response of the chart with the colourspace. Just softproof.

2)For canned profiles available from the various paper manufacturers this chart is really handy. You see when you make a profile you can specify the rendering intent. Experimenting with softproofing can lead you to discover which rendering intent the profile was made for. For example using an innova ifa49 profile, Rel introduces a colour cast whilst per stays neutral.

3) softproofing with the chart & a canned or custom made profile lets you know how well the profile was made & how to compensate for it when printing an image. For instance in the example above with innova and ifa49 the profile (either in rel or per rendering) wont show shadow detail any lower than about 5/6. Apply a curves adjustment layer to the image whilst softproofing and click on the node in the bottom left (representing 0 or black) and change in the input/output value from 0/0 to say 0/5. This lightens the test chart allowing the shadow detail down to 1 to be seen. This can then be fine tuned and the adjustment layer dragged over to your image for printing. You can then print with this adjustment layer applied, and the correct rendering intent as discovered by softproofing the test chart.

4) The Epson r2400 profiles probably don't offer black ink simulation, because they were made by a complete idiot who saved them as a display (ie monitor) or input device profile. Due to this they can't simulate black ink because photoshop doesn't think they are paper. The printer driver obviously ignores this, but, it gives you an idea of how crap they are......

5)I'm looking at your screen grabs on an Eizo CG241 monitor. In the grabs, I can see between 1-2 more stops more highlight than you've mentioned in your numbers. I can also see farther into the shadows. On the ILFORD SMOOTH PEARL PAPER I can see down to 6 on perceptual & 9 on relative. Those Harman profiles are interesting. See how all of them have the same issue around the 249-251 mark. This would have been down to a possible error at the profile making stage. Again you could apply a curves adjustment layer and sample the area which isn't resolved 252 upwards and pull the curve down to bring it within the profiles range. You can then drag this curve over to your image in P-shop at the printing stage.

To summarise these profiles aren't bad. You got to remember that they are designed for printing straight to the printer through the standard windows or mac driver. With a rip it would be a different kettle of fish with all points on the chart resolved.

 

[Marco B]

Hi Scovell001?

Thanks for stepping in and taking my observations to the next level including some suggestions to deal with the issues, like the curves adjustments.

Its a really bloody useful image and I now use it as part of my profile testing all the time.

Yes, I agree... it is a very basic test image, but it has definitely thought me some new things.

1) don't assign or convert the image to Adobe RGB before you start soft proofing, as you are adjusting the tonal response of the chart with the colourspace. Just softproof.

Yes and no, I can understand where you are going with this, but since most images already have a predefined sRGB or AdobeRGB assigned, the proofing of an unassigned image won't tell you much about how the image with assigned colorspace will react / print, because that one will have the "adjusted tonal response" as you described it. You might fool yourself into thinking you will get more shadow separation than the actual photo with (camera) assigned AdobeRGB will accomplish...

Or am I missing something here :confused: Please do respond to this one!

2)For canned profiles available from the various paper manufacturers this chart is really handy. You see when you make a profile you can specify the rendering intent. Experimenting with softproofing can lead you to discover which rendering intent the profile was made for. For example using an innova ifa49 profile, Rel introduces a colour cast whilst per stays neutral.

Actually, this is a good point, why don't manufacturers document what rendering intent they used to create the profile or, maybe better, like you said "the rendering intent it was made for"? :confused: As they must have used one to create the profile in the first place.

3) softproofing with the chart & a canned or custom made profile lets you know how well the profile was made & how to compensate for it when printing an image. For instance in the example above with innova and ifa49 the profile (either in rel or per rendering) wont show shadow detail any lower than about 5/6. Apply a curves adjustment layer to the image whilst softproofing and click on the node in the bottom left (representing 0 or black) and change in the input/output value from 0/0 to say 0/5. This lightens the test chart allowing the shadow detail down to 1 to be seen.

Good suggestion: for those reading this: just be aware that PS can have the curves black / white switched in the tools window. So setting an output value of 5 might actually do the wrong thing if the windows black / white is swapped (it was when I tested it compared to Scovell001's remark).

There is just one drawback maybe: using this method, you lighten up the whole image. If you have the bit depth (16 bit / channel), a specific ligthening up of the shadows might be better...

4) The Epson r2400 profiles probably don't offer black ink simulation, because they were made by a complete idiot who saved them as a display (ie monitor) or input device profile. Due to this they can't simulate black ink because photoshop doesn't think they are paper. The printer driver obviously ignores this, but, it gives you an idea of how crap they are......

My thought too, but I figured it would be wise to keep it polite until someone like you could confirm my observations.

Actually, it is quite shocking for a well known printer manufacturer like Epson to provide such profiles... :rolleyes:

5)I'm looking at your screen grabs on an Eizo CG241 monitor. In the grabs, I can see between 1-2 more stops more highlight than you've mentioned in your numbers. I can also see farther into the shadows.

I am not surprised by your observations. I don't have a top notch monitor at this moment like your Eizo. So, although it is properly calibrated and does a reasonable job, it won't show it all. But I already observed this since I noticed the actual prints showed a little bit more too, in accordance with your observation. See my remark about this in the previous posts.

Actually, I was pleasantly surprised to see how close the Harman Matte Warmtone baryta paper was to the soft-proof, a sign that this specific factory supplied profile was quite acceptable, as is the Photoshop's soft proof rendering. Although there are clear limitations though comparing results from such different media as a screen and print...

Marco

 

[toyotadesigner]

The effect with the matte paper is obviously identical to darkroom or photolab prints: if you use a paper with a glossy surface you'll get more details in the shadows and an overall better tonal range.

AFAIK it's because the glossy media reflect more light, which is particularly important for the shadows or dark areas: the more light will be reflected the more details will be visible. The human eye sees less contrast in low light (sorry if I can't explain this effect in more detail but I don't know all the expressions for it in English :confused: ). The spectrometer issues light and measures the reflected rays. So, because of the nature of matte paper the eye as well as the spectrometer can't differentiate as many details as compared to glossy paper.

Other than that I can't help too much because I have a HP Pro 9180B which delivers excellent prints without too much input from my end.

Sure, there is a difference in the tonal range between glossy and matte or canvas, but knowing this I always print color images with glossy paper, b&w with matte paper or canvas, using the effect to produce a deeper black with b&w.

Damn it, I just notice how bad my English turned over the last months - a bad side effect of living in the deep south of Spain with little contacts to the rest of the world.