Q re. shadow & highlight points

[mikemeister_admin]

This may be a CM newby Q, forgive me.

Do I understand correctly that the shadow & highlight points set those portions of the image to the very edges of the image tonal range once the CM modifications are made and the user returns to the PS interface?

If so, then it means that, for example, a point in the image that measures RGB 20, and which gets used for the CM shadow point, would get set to RGB 0 and a point with RGB 240, and which gets used for the CM highlight point, would get set to RGB 255?

Hmm, I usually set my tonal range endpoints to RGB luminosity 15 & 245 respectfully (for printing purposes). Is there a way to have CM respect these endpoint values when using the CM shadow & highlight points?

Note: in my current workflow I set the tonal endpoints using a Threshold adjustment layer to find the darkest and lightest points (that are of interest and also not reflections) in the image and then a curves layer using the black and white point eyedroppers.

[-default]

They are set to extreme values by default.  You can change the shadow and highlight points to any value in the settings (these are special in that they are used for calculating the threshold display), or create a shadow and highlight pin manually to enforce any values that you want.

[mikemeister_admin]

Note: in my current workflow I set the tonal endpoints using a Threshold adjustment layer to find the darkest and lightest points (that are of interest and also not reflections) in the image and then a curves layer using the black and white point eyedroppers.

Howie, I really like the CM threshold locating method. If you move your cursor to the left of the RGB curve window, right against the border, it will turn to a double line (shadow threshold) and when you click on it, will show you the shadow threshold as promised. Likewise the highlights. What I like best is you can still see the image clearly to help set your points. I normally just hunt around with an eyedropper and watch the numbers, but this is a very easy alternative.

Greg's been keeping this pretty close to his chest, though. Pried it out of him just yesterday.

[mikemeister_admin]

Yes, I also like the way CM allows the user to find the darkest and lightest portions of the image, and, as you point out, it gives the luminosity value for those darkest points. The issue is that if the user uses the shadow & highlight tools, or the double line thresholds (http://curvemeister.com/support/curvemeister3/help/Examples/04-UseThresholds.htm), the digital tonal range of the image will extend from RGB 1 to 255 once the CM adjustments are made and the user returns to the PS interface.

Try this with an image that has a short tonal range:
1. Open the image in PS and then CM.
2. Set the color space to Lab.
3. Use the thresholds to find the darkest and lightest points in the image. Let's say that the darkest point is L75, and the lightest point L25.
4. Apply these settings and return to the PS interface.
5. Re-open CM.
6. Now check the thresholds. You'll find that the darkest point is L100, and the lightest point is L0.
7. Change to the RGB color space.
8. Check the thresholds. You'll find that the darkest point is 1 and the lightest point is 255. As you know G, if you print this image everything from RGB 1 to ~15 will be black without any detail since the printer can't replicate those tonal variations. Likewise, everything in the range RGB ~245 to 255 will be the color of the paper without any tonal variation since there is no paper that is 255. (Note: depending on the paper used there may be variation up to ~248 but it still means that there are hightlight variations that effectively get clipped because the tonal range has been digitaly set to the complete RGB 1-255 (L100-0) range.

So, what I'm wondering is if CM can allow the user to use thresholds and/or the shadow and hightlight tools but set those darkest and lightest points--once the CM adjustments are applied--to a user defined limit (i.e. RGB15-245 // L97-2) rather than force those darkest and lightest points to RGB1-255 // L100-0).

Ok, one might point out that the user could, once the darkest and lightest points are found, back off the values used. In my example it would mean manually setting the darkest point to L78 and lightest point to L22. But it would be more user friendly to allow the user to choose clipping values in the CM settings.

What's more, there is, AFAIK, no discussion in the wiki or help manual, save for the short note in the help manual about low & high key images, that points out the danger of using the whole digital tonal range and that it causes clipping in the print's tonal range.

[-default]

You are very correct that this could be better documented.  For those concerned about such things, considerable attention has been given to the issues that you mention, going back back to Curvemeister 1.  There are settings for the shadow and highlight in the prefs - these are the values used to calculate the threshold image.  You can also create your own shadow and highlight pins for a variety of viewing and/or press conditions.

This, combined with corner fill's gradient fill, provides a convenient way to get exactly the tonal range that you want - I think more precisely than is easily obtainable in Photoshop.

[mikemeister_admin]

There are settings for the shadow and highlight in the prefs - these are the values used to calculate the threshold image.

Are these the RGB Shadow and Highlight Targets found in Settings?
If so, then yes, I agree, this could be better documented. It would have surprised me if this functionality were not already part of CM.

They seem to be working as expected (shadow point being set to my desired 15,15,15 and highlight point being set to 245,245,245). Although I don't seem to be able to get the desired precision with the shadow pin. Maybe it's just the image I'm currently using.


Edit:
Hmm, now I'm not so sure. The Image Window Options page says that my custom targets will not be used if I use pins.

This shadow value is overridden if a pin is used as a shadow, and that pin is used as the new default for subsequent shadow points invoked via a right click on the image window.

Does this include the shadow and highlight pins? If so then I don't understand the point of the shadow and highlight targets.

[-default]

This shadow value is overridden if a pin is used as a shadow, and that pin is used as the new default for subsequent shadow points invoked via a right click on the image window.

> Does this include the shadow and highlight pins? If so then I don't understand the point of the shadow and highlight targets.

This used to be true, but is no longer.  (BTW, while doing one of your other changes, I changed the manual to reflect this)

The history of this is that in Curvemeister 1 there were no color pins, ao the shadow/highlight targets were provided to change these values.  When color pins were added in Curvemeister 2, I thought it would be reasonable to temporarily override the target values with values from the last shadow or highlight pin that was used. 

When looking over this code for Curvemeister 3, I decided that the above behavior was too unpredictable, and Curvemeister 3 now always uses the configured target values for calculating which areas of the threshold image should appear black and white, and for the built-in shadow, highlight, and threshold functionality.

You might also be interested to know a couple more design considerations that touch on shadow/highlight functionality.  The first one is that corner fill plays an important role in "filling out" the black and white areas that fall outside of the range of the shadow and highlight.  For example, setting a shadow point of RGB(10,10,10) will offset the curve point vertically.  Corner fill will "fill" the area from the shadow to the origin with a linear gradient, rather than clipping it to RGB(10,10,10).  Similar logic holds for the highlight.

The second detail has to do with wgCMYK, and even CMYK.  The logic for setting the highlight is about the same as for the other color spaces.  OTOH, setting the shadow point, and calculating the threshold image, had some surprises.  In particular, although the K value is generally high in the shadow, the other three color channels, depending on black generation, may be quite light.  Curvmeister deals with this by adjusting the K value via its endpoint, and confining its adjustments to the CMY channels to interior curve points.  This behavior can be seen when you drag a shadow point across a neutral gradient, for example.  I think this solution works pretty well.

Thresholding the preview image shadow is another matter.  At the moment, I simply punt when calculating the threshold image for these color spaces - the highlight behaves as a normal threshold, but the shadow is not thresholded at all.  For wgCMYK, it's pretty easy to fix the problem by thresholding the sum of each channel with the K channel (and I plan to add this at some point).  For true CMYK, the answer is not obvious at all.  What would the characteristics of a thresholded CMYK shadow preview be?  I'd appreciate any comments you (or anyone else who made it to the end of this post!) might have.

[mikemeister_admin]

...corner fill plays an important role in "filling out" the black and white areas that fall outside of the range of the shadow and highlight.  For example, setting a shadow point of RGB(10,10,10) will offset the curve point vertically.  Corner fill will "fill" the area from the shadow to the origin with a linear gradient, rather than clipping it to RGB(10,10,10).  Similar logic holds for the highlight.

From Curve Options: "Corner fill improves the behavior of CMYK black points, and is also useful for bringing out shadow and highlight detail since it approximates the "lizard tail" curve shape.. "

Corner fill may explain the behavior I was seeing when testing the use of shadow targets. Would you mind explaining how corner fill improves the behavior, and why this is only applicable in CMYK?

Thresholding the preview image shadow is another matter.  At the moment, I simply punt when calculating the threshold image for these color spaces - the highlight behaves as a normal threshold, but the shadow is not thresholded at all.

...or does this explain why my tests showed that highlight targets and thresholding was quite accurate but shadow targets and thresholding were not at all as precise?

[mikemeister_admin]

I haven't looked carefully at the s/h eyedroppers--simply because I shoot raw images, and they come into CM with tonal range already carefully determined.
But Howie got me thinking, and since there are times I will have other previously developed files to work on, I thought I would check out the s/h options more carefully. I changed my settings to my preferences--10 and 247. I then tried both ways of setting the highlights. Using the highlight tool on the toolbar, and selecting a point in the rightclick menu. Both worked. Then I tried the shadows. I am getting very inconsistent results. Usually I get 0,0,0 readings (I'm in RGB). Sometimes 3-5 range. Occasionally 10,10,10, but rarely. At first it seemed that if I deleted the point and then tried again the results were always 0,0,0, and I had better results if I used the reset button in the curve window. After many, many tests, though, it just seems inconsistent. I tried zooming in to make sure I wasn't shifting the eyedropper too much, but no improvement.

Howie seems to be getting the same results:

...or does this explain why my tests showed that highlight targets and thresholding was quite accurate but shadow targets and thresholding were not at all as precise?

If corner fill is the culprit, I'll eat my hat. :)

Gloria

...I'm starting to get a little worried about my hat...

[mikemeister_admin]

G, just because we're shooting raw and using ACR or LR for conversion doesn't mean that we always nail the shadow threshold target--at least I don't. As you already know, I've learned to be far more conservative in setting my tonal range endpoints in LR: even still I'm finding that I don't always make my mark of 15 and 245. Often my shadows are too dark.