Rendering Intent

I have an HP 9180 printer and I use Qimage Studio to manage the colors for printing my Abobe RGB colorspace managed tifs. I have been printing using the perceptual rendering intent from Qimage's color mgmnt tab...it's the default. My prints seem to be great and closely match my image viewed on the monitor.

From reading the many pages on the subject (rendering intent), it seems that most of the time when using photoshop to manage the colors for printing, relative colorimetric is chosen with the black point option checked. Should I be using this same rendering intent option when using Qimage for printing ? What is th advantage if any. I could always try a print with this option, but I was looking for some guidence first. TIA.

c.w. moynihan wrote:I have an HP 9180 printer and I use Qimage Studio to manage the colors for printing my Abobe RGB colorspace managed tifs. I have been printing using the perceptual rendering intent from Qimage's color mgmnt tab...it's the default. My prints seem to be great and closely match my image viewed on the monitor.

From reading the many pages on the subject (rendering intent), it seems that most of the time when using photoshop to manage the colors for printing, relative colorimetric is chosen with the black point option checked. Should I be using this same rendering intent option when using Qimage for printing ? What is th advantage if any. I could always try a print with this option, but I was looking for some guidence first. TIA.

Christian,

I typically find that it is best to soft proof in PS before printing to select the best rendering intent for that particular image. Some settings deal with certain out of gamut colors better than others. I even have a few images that print better with Black Point Compensation unchecked. Most often however I prefer Relative Colorimetric with Black Point Compensation checked. That's kind of my default setting.

I almost always use Relative Colorimetric with BPC when printing from Qimage. The reason is that I maintain relatively close control of my image color through post-processing. Relative Colorimetric preserves all in-gamut colors during transformation into the target color space (with the caveat that they are converted relative to the target white point), only clipping any colors that are out of gamut. Whereas Perceptual will compress colors that are in as well as out of gamut if any colors exceed the target gamut, in order to preserve the relative relationship of all colors at the possible cost of introducing some color shifts.

I personally don't want color shifting to go on unless I'm controlling it, so RC is my normal choice. If for some reason I had an image containing a bunch of out of gamut colors and couldn't manually address them during post-processing, then I might use Perceptual to avoid losing detail in saturated colors when the out of gamut ones were clipped by RC. But my ideal solution given the time to do so would be to manually adjust the image color to alter the out of gamut colors, and then print via RC.

Thanks Les and Royce. Much appreciated.

Depending on which printer profile you're using, you may find that the Perceptual intent provides somewhat better shadow detail than the Relative Colorimetric intent (even when Black Point Compensation is enabled). This is am important consideration for images that contain significant shadow detail.

Eric, do you know the basis of the effect of Perceptual on shadow detail? I've infrequently seen it mentioned but never attempted to track it down and understand when & why it's going on...

One other thing occurs to me. If a person does like I do and maintains fairly close control over colors during post-processing to eliminate any that are out of gamut, properly handle saturated detail & smooth tonal gradients, etc., then the gamut clipping / shifting behavior of the different rendering intents should be a moot issue. There are no out of gamut colors to trigger clipping or shifting. In that case, then use of Perceptual or Relative Colorimetric should be basically equivalent in terms of the accuracy of colors.

There may be a different reason (or reasons) to use one RI or the other, though, such as the shadow detail point that Eric has brought up.

Whats the best way to look at the image to see if colors are out of gamet ? I guess this is where I need to do a soft proof in PS ? I wonder if I can soft proof and detect out of gamet issues in Qimage ?

Thanks again. Great information here.

howlandr wrote:One other thing occurs to me. If a person does like I do and maintains fairly close control over colors during post-processing to eliminate any that are out of gamut, properly handle saturated detail & smooth tonal gradients, etc., then the gamut clipping / shifting behavior of the different rendering intents should be a moot issue. There are no out of gamut colors to trigger clipping or shifting.

But presumably the colour control to which you refer is in your editing space. Is it completely contained within all of the print spaces you use? Or are you saying you manually tailor the image to each print space?

Christian -- yes the best way to soft proof is in Photoshop. Qimage has a soft proof function but it's pretty rudimentary. In particular it doesn't support a gamut warning ability like Photoshop has. If you use Photoshop to do your post-processing edits and Qimage to do your printing, just soft proof in PS as the final stage check before you jump across to Q. You'd have to come back to PS anyway if soft proofing showed any issues in the image, because most likely you wouldn't be able to to effectively address the situation using Q's color controls.

Stephen -- the color control I'm referring to is making a set of output-specific adjustments to the image while soft proofing it using the intended output profile. While the image and PS working color spaces usually are both ProPhoto RGB in my workflow, I'm making adjustments designed to tune color, contrast, etc. for the print profile. So it is the latter -- I am manually tailoring the image to whatever output color space I will be using for output. If I later decide to print using a fairly different output profile, I double check those print adjustments and redo them as necessary.

howlandr wrote:Christian -- yes the best way to soft proof is in Photoshop. Qimage has a soft proof function but it's pretty rudimentary. In particular it doesn't support a gamut warning ability like Photoshop has. If you use Photoshop to do your post-processing edits and Qimage to do your printing, just soft proof in PS as the final stage check before you jump across to Q. You'd have to come back to PS anyway if soft proofing showed any issues in the image, because most likely you wouldn't be able to to effectively address the situation using Q's color controls.

Makes total sense Royce. As always, thank you very much for your technical assistance.

Printing is fun !!!

Hi Royce,

The issue with shadow (and to a lesser extent, highlight) detail when printing is as follows. Each ink/paper combination has a white point and black point ... you can think of these as maximum and minimum reflectivity. As we know, no paper can reflect all light, and no paper can absorb all light. So the reflective properties determine the contrast range of the paper (e.g., L* = 6 to L* = 95 for an average luster paper on modern inks).

Now, what happens if we try to take an image which has a full contrast range from RGB = (0,0,0) to RGB = (255,255,255) (i.e., L* = 0 and L* = 100, respectively) and print it on this paper? Somehow the image values have to be remapped to fit within the contrast range of the paper. This much I'm sure you understand already, but just laying it out all here so we have all the context ...

The Relative Colorimetric intent specifies how the white point of the image will be remapped to the white point of the paper, but says nothing about how the black point will be mapped. When Black Point Compensation is disabled, clipping occurs and we sometimes will end up with a horrible loss of shadow detail. With Black Point Compensation enabled, the BPC algorithms make an educated guess about the black point of the paper and then try to compensate accordingly to preserve shadow detail.

In contrast, the Perceptual rendering intent -- being a secret sauce of the profile-building software -- is allowed to do whatever it wants, and consequently most Perceptual tables are built in a way that take into account both the white point and the black point of the paper. That is to say, there is less guesswork involved here than with the BPC algorithms, since the black point information is determined directly from the profile target's measured data (i.e., what is the deepest black obtainable on this paper?). Consequently, a well-built perceptual table will therefore have mechanisms to preserve shadow detail, based on the notion that the exact black point of the paper is known.

Now, to be honest here, and to return this conversation to plain English, it may be the case that the differences between the Perceptual intent and the Relative Colorimetric intent in terms of shadow detail are subtle. In some situations you may see no difference at all. It does depend on which profile you have, since it depends on which software (and software options) was used to build that profile in the first place.

I would also like to caution everyone that Photoshop's out of gamut warning tool is just an estimate. I have seen cases where there are colors in an image that are definitely out of gamut, but do not show up in the Photoshop window when the OOG warning is enabled. So, use it as a guide, but perhaps pay more attention to the actual Soft Proof appearance than to just the warning color.

Eric, thanks. I guess I've been operating under the assumption that BPC was more deterministic & useful than you are describing since the black point is part of the measured response of the media. Sounds like BPC is a hack. I've believed that Perceptual was more "secret sauce" driven and thus less deterministic, i.e. unpredictable. Sounds like in some cases Perceptual's very lack of determinism may be a benefit in that it can be "cooked" to produce a better result due to the inherent flexibility of how it handles the mapping. Hmmm.

But your point raises something else that I've been doing in my workflow, which I learned from Darwin Wiggett. I never send L* = 0 or L* = 100 to the printer these days. In most cases I manually set the black & white points in the image to achieve black point of about L* = 5 and white point of about L* = 95. This is one detail within my blanket statement earlier about how I attempt to exercise fairly careful control of the color, luminance, etc. of my images. So in the end I may be winding up in the same (or similar) place but with manual controls in place of secret sauce.

Interesting, Royce. Why are you sending L* = 5 and L* = 95 to the printer? Does this mean that if you actually have a patch of solid black that you wish to reproduce in the print, that instead of sending RGB = (0,0,0) (which is L* = 0) you actually change the image value to RGB = (14,14,14) (which is L* = 5 for ProPhoto RGB)? And that instead of sending RGB = (255,255,255), which is L* = 100, you actually change the image value to RGB = (237,237,237)?

I think I must be misinterpreting what you're saying, because if you actually performed the steps above, you'd be introducing a fairly drastic reduction in the overall dynamic range of the print ... !

Yes, you understood what I was saying. I wasn't sure I initially understood what Darwin was saying, and had to ask for clarification. While Darwin initially suggested L*=5 and 95, I just checked my settings and I've tweaked it slightly to black point of L*=4 (10,10,10) and white point of L*=96 (245,245,245).

I should not have said "never send"... never say never. But with my HDR work especially, there may be no L*=0 in the frame in any event. If a scene doesn't contain pure 0,0,0 black I'm not a fan of introducing it through a brute force levels shift just to get something that uses a max amount of the histogram & lays down as much black ink as possible. In fact it's the opposite. Because of the shadow detail issue we've been talking about, I sometimes find that even if the image naturally contains L*=0, printing this as-is can result in less than optimal shadow detail. I suppose it's partly a shadow curve compression issue, and partly an issue with the printer's (or profile's) ability to render dark tones with detail since my default rendering intent is Relative Colorimetric with BPC.

On the white side I do back things down as described. My interest in doing this is to avoid completely blank areas of paper showing through, which will be present with any L*=100 in the image. These spots introduce a different white cast in the highlights because paper white usually doesn't match image white. On the R2400, paper showing through also causes gloss differential which I don't like.

Photomatix has black & white clip settings that can be applied during tone mapping, and I used to set them to very slightly clip on both shadow & highlight ends to make sure I was getting as much of the histogram populated as possible during the compression from the HDR image data down to a 16-bit TIFF. But I eventually decided that this was actually costing me loss of detail at the extremities. Also the possibility for the odd artifact on the blackest tones where conversion from HDR's 32-bit floating point luminance values into 16-bit integer values may have some precision issues. So now I do not clip at all in Photomatix tone mapping, and it may be likely that 0,0,0 and 255,255,255 do not occur as a result.

Around the time I was fiddling with this stuff and working some more on my print workflow, I went on the workshop and Darwin mentioned the trick of setting the black & white points slightly in from the edges. I did some further checking into it and it has been written about by numerous folks as a technique for getting the darkest & brightest tones to hold more detail in print. Technically it does reduce the print's dynamic range. Pragmatically it seems not to be a big hit, with benefits that are worth it. It's another one of those "last few percent" quality things that I seem to be spending a lot of time digging into.

Among the numerous bits of info on this technique, see this workflow from Michael Reichmann; scroll down to find the section "6. Set Print Map Target Preference":
http://www.luminous-landscape.com/tutor ... low1.shtml

The technique derives originally from offset printing on coated stock where best results involved sending a 5% dot for white and a 95% dot for black. The people who advocate the technique feel it translates to inkjet printing as well.