Dan Margulis Applied Color Theory

The "Jacob's Ladder" Technique

L* channel duplicated across a* and b* for finer color control, &c.
Posted by: Jacob Rus
Tue Sep 2, 2008 2:42 am (PDT)

Hi List,

This has possibly been discussed before (I haven't run any exhaustive searches of the list archives, and haven't more than skimmed Dan's book in the bookstore, so let me know if I'm saying something well known) but I haven't ever seen it mentioned in other books or in the various color-correction internet material I've seen, so here goes.

For problematic images--i.e. those with differing color casts in highlights, midtones, and shadows, or possibly casts appearing in only particular hues--or for artistic effect, the blend if sliders (or alternately various sorts of masks) are useful for limiting the range of various curves corrections, etc. But I have stopped going to them first for complex corrections in these cases, in favor of the following, which is sort of my way of hacking Photoshop into a color correction GUI I find more intuitive than using the Curves tool "straight":

* Convert the image to CIELAB (obviously)

* Copy the L* channel (I

* Create a new layer, and paste the L* channel into all three channels (L*, a*, and b*).

* Set the blend mode of this new layer to "Linear Light,"[†] and bear with me here, because at this point the image will look like a black/white posterized splotch. I usually name this layer something like "L*-based corrections".

* Add a new curves adjustment layer, and set it as a clipping mask.

* Make all three curves--L*, a*, and b*--completely flat, i.e. Input 0 -> Output 50, In 100 -> Out 50 in the L* channel, and In -128 -> Out 0, In 127 -> Out 0 in a* and b*. At this point you should have an image that looks exactly like the original, because Linear Light blend mode makes no change to the bottom layer where the blend layer is middle gray.

* Now here's the fun part: make some points along those curves and slide them up and down. You can thus adjust (linearly shift) L*, a*, and b* values for various background levels of L*, with much finer control than using the blend if sliders, or any other tool I know.

- Want more shadow contrast? Leave the point at i0->o0, and add another couple at, e.g., i6->o53 and i25->o54.
- Shadows a bit too green? Make an upward bump in the left end of the a* curve.
- Sky not blue enough? Make a downward bump in the right end of the b* curve.

&c.

That's it. If you want to make corrections based on the a* or b* values, instead of based on L*, the procedure is identical, except copy the a* channel (or whatever) into each one of the new layer's channels.

The main down-side to this approach for me is that Photoshop's Curves controls here have a very big range, and the corrections to be made with this tool tend to be *very* close to that center line: I sometimes wish I could zoom the curve's vertical axis to only show its middle third (or in the case of adjusting based on underlying a* and b*, zoom both axis like that).

Anyway, I realize this email was a bit long, but hopefully others find this tool as useful as I have. It becomes especially nice after setting up an action to create these layers, but making them manually isn't *too* arduous.

Cheers,
Jacob Rus

[†]: A few other blending modes also work, but have nearly the same possibilities here, and linear light is easier to use in my experience. But feel free to experiment :-)

P.S. This is my first posting to this list, so I should probably introduce myself. I'm a Harvard undergraduate government concentrator--formerly in the math department, but I got tired of problem sets, took a year and a half off to work on various programming stuff, and am now back for the last year.5 of school. I have been a huge Photoshop fan for about half my life now, and I try to get out to take pictures as often as I can with my D50 (never often enough). I continually tell myself I need to put some up online, but I somehow manage to never get around to registering a domain and coding up the simple site I want. Sometime soon though, I promise!

P.P.S. Can anyone recommend the best introductory/intermediate book about image processing? I just this week started trying to learn how to use NumPy (array math for Python) so I can try to build some of my own types of tools, mostly for own edification, but also because I sometimes find Photoshop's color correction tools limiting in ways that seem easy enough to improve on, but I want to "put my code where my mouth is" so to speak. :-)
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Tue Sep 2, 2008 4:49 am (PDT)

Okay, a few follow-ups (which are left here because I forgot to mention a couple of them, a couple are more detailed than needed for the initial message and would possibly impede comprehension, and I just figured one out right now).

1. This approach (copying one channel into all of the channels, setting blend mode to Linear Light, and then adjusting a flat curve) can also be used in RGB and CMYK to good effect. In those color modes, I basically use it as a more powerful channel mixer + blend if combined. If you've been unsatisfied with the power available in channel mixer, I recommend you play with this, for example for adding back detail to very colorful objects.

2. This is a way to mix L*/C/M/Y/K channels into RGB images, and vice-versa. Just duplicate the image, convert to the new space, copy the desired channel, paste it back into all channels of a new layer set to Linear Light, add the clipping adjustment layer, and adjust to heart's content.

3. For something even potentially cooler, try:

- Duplicate the image
- Convert the duplicate to CIELAB
- Use the Hue/Saturation tool to rotate the hue 30°[†]
- Convert the image back to RGB or CMYK
- Copy one of the resulting channels
- Go back to original image
- Paste into all three channels of a Linear Light layer
- &c.

This lets you mix channels which are half-way between RGB and CMY (roughly: modulo the differing luminosity contributions of different wavelengths), which means you can alter colors along sometimes more convenient axes than RGBCMY.

4. Various channel mixing uses of this technique can be very effective for converting a color image to grayscale. Indeed, the combination of this technique, the Shadow/Highlight tool, and various sharpening/blurring of various channels on large/small scale, is enough to handle pretty much all my B&W conversion needs. :)

5. Once you have a B&W image, the technique can also be an effective tool for tinting it with extreme control. Just (in CIELAB mode) set up a layer with L* copied to all three channels (no need to set this to Linear Light this time), and then create a curves adjustment layer, with the L* curve left as a normal ramp from i0->o0 to i100->o100, and then set the initial curves of the a* and b* channels to be flat 0 all the way across. Then move a* and b* curves up and down to tint various L* levels of the image different tints, without affecting the L* component of the resulting colors.

6. I said before:

The main down-side to this approach for me is that
Photoshop's Curves controls here have a very big range,
and the corrections to be made with this tool tend to
be *very* close to that center line: I sometimes wish
I could zoom the curve's vertical axis to only show its
middle third (or in the case of adjusting based on
underlying a* and b*, zoom both axis like that).

I just figured out how this can be solved. Woohoo! We can make a Levels adjustment layer between the 3-alike-channels layer, and the nearly-flat-curves-meat-of-the-technique layer, and use it to "zoom" the horizontal axis, by bringing in the input left/right sliders by the desired factor (if we're trying to edit based on a* and b* values, they are often very low, so we can bring compress the levels here to 1/5 or even 1/10 of their original extent. This lets us use more of the curve in our main curves layer.

Likewise, to "zoom" on the vertical axis, we can add a Levels adjustment layer *above* our main curves layer (note all of these layers should be set as clipping masks), and bring its *output* sliders (on both ends of every channel) in to 1/5 or whatever their original extent, which will effectively dampen whatever adjustments we make in our main curves layer, which means we can see the adjustments we're making better, because they won't all be so close to the axis.

I'm excited about this. It makes using this technique much faster because less precise mousing is needed.

Okay, enough for now. I'll give others a chance to weigh in (e.g. to tell me they think I'm completely nuts). :-)

Cheers,
Jacob Rus
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: "Werner Tschan"
Tue Sep 2, 2008 4:49 am (PDT)

Dear Jacob

I tried your move. What I didn't get was a b/w posterized image but a colored one. I chose a photographic reproduction of a painting of a New York street scenery to try it on. In one of the corners the tarmac is turning from slight blue into a very saturated blue. In the open curves dialogue I picked up the blue in the b channel, and lifted the curve just at this point and sure enough the blue turned neutral without affecting any other parts of the image. What a fantastic work around! Thanks for this most valuable input. I am waiting for more!

Werner Tschan, Switzerland

----------------------------------------------------------
STUDIO LTD
Werner Tschan
Atelier für Fotografie
Altenbergstrasse 8
CH-3013 Bern

T +41 (0)31 332 88 33
M +41 (0)78 227 02 19
U www.studio-ltd.com
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Tue Sep 2, 2008 4:49 am (PDT)

Jacob Rus wrote:

[...] Use the Hue/Saturation tool to rotate the hue 30°[†]

Oh, the forgotten note that went with this was:

[†]: You experts all probably already know this, but the Hue/Saturation tool is *way* more useful in CIELAB mode than in RGB mode. (It modifies the image in CIELAB LCh, rather than the brain-dead HSL)

-Jacob Rus
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Tue Sep 2, 2008 6:49 am (PDT)

Werner Tschan wrote:

I tried your move. What I didn't get was a b/w posterized
image but a colored one.

Erm, you're right, of course. Not sure what I was thinking when I wrote that line. :-)

-Jacob
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Stephen Marsh
Tue Sep 2, 2008 6:49 am (PDT)

Thanks Jacob! Most of the negative issues associated with RGB mode Hue/Saturation command tweaks in Photoshop can be addressed by:

a) Setting the adjustment layer or fade command to Color or Hue or Saturation blend mode

b) Working in a wide gamut RGB space (even if for a temp saturation boost move then back to a smaller space)

Sincerely,

Stephen Marsh
___________________________________________________________________________

PS's Hue/Saturation in RGB vs CIELAB (was: Re: L* channel duplicated
Posted by: Jacob Rus
Tue Sep 2, 2008 10:54 am (PDT)

Thanks for the advice. I do use Hue/Saturation sliders in RGB very infrequently, for tweaks to nearly solid-color objects, which I can isolate in a narrow hue range, without affecting the rest of the image. If these tweaks work for you, and give the kind of results you are looking for, great! I certainly won't argue with experience. Also, my wording wasn't meant to be inflammatory. I should have said "*way* more perceptually accurate" rather than "*way* more useful".

(And this discussion is also sort of off topic: I really would like most to hear what people think of my flat curves in Linear Light blend mode idea. :-)

But from a technical standpoint, any Hue/Saturation manipulation using HSL, no matter the particular RGB/CMY space it's based on, is going to:

* Alter lightness (as in L*) while it is ostensibly modifying chromaticity (H/S) alone.

* Alter chroma (as in Munsell/CIELAB LCh/CIECAM02) while ostensibly modifying only hue, and alter chroma relationships while increasing or decreasing "saturation" (that is, boost chroma in some colors more than others, in an unpredictable non-uniform, non-linear way).

* Rotate differing hues by different perceptual amounts, altering hue relationships between colors.

You can (somewhat) get around the first of these by using a blend mode. But that leaves the other two, which are as far as I'm concerned quite serious. If you're altering photographs, and sort of roughly eye-balling things--tweaking until they look good--then this perhaps doesn't matter so much, particularly if you stick to making small tweaks to limited parts of the hue range.

The real problem is that HSL and HSV were color transformations devised in the 1970s to be very computationally efficient on 1970s-era hardware (i.e. could be done on-the-fly for moving interactive graphics, &c.), rather than designed for any particular resemblance to human perception.

This is what causes Charles Poynton to [say][1]:

HSB and HLS were developed to specify numerical Hue, Saturation
and Brightness (or Hue, Lightness and Saturation) in an age when
users had to specify colours numerically. The usual formulations
of HSB and HLS are flawed with respect to the properties of colour
vision.

[... and he goes on to bash them some more ...]

And [here][2]:

[...] For these reasons, any use in computer graphics of I, B, L,
and V quantities is suspect.

[1]: <http: //www.poynton.com/notes/colour_and_gamma/ColorFAQ.html#RTFToC36>
[2]: <http: //www.poynton.com/notes/colour_and_gamma/GammaFAQ.html#HSI>

If you care strongly about retaining color relationships while altering hue/chroma, there is little reason not to just use the Hue/Saturation tool while in CIELAB mode. (And I'm sorry that there isn't a "UPLab" [Lindbloom's ICC profile which passes CIELAB through a LUT based on the Munsell data] or CIECAM02 mode--either would be hot, as CIELAB has some nasty curvature of perceptually constant hues in the blue/purple region.)

But this is almost surely a re-hash of previous discussions on this list? Anyway, I'm not super interested in pursuing it.

Cheers,
Jacob Rus
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: "chris broadhurst"
Wed Sep 3, 2008 2:19 am (PDT)

 Jacob Rus wrote:

This has possibly been discussed before (I haven't run any
exhaustive searches of the list archives, and haven't more than
skimmed Dan's book in the bookstore, so let me know if I'm saying
something well known) but I haven't ever seen it mentioned in
other books or in the various color-correction internet material
I've seen, so here goes.

Jacob,

I think this is one of the most innovative and powerful ways to colour correct an image that I have heard about - it was an instant WOW when I tried it, and it's concept has opened up some new ideas for me to experiment with. The only weird thing was curving from a horizontal line!

Thank you so very much for sharing it with us - I trust you dont mind, but I've spread your idea around the CurveMeister forum.

Chris Broadhurst
Web Site: http://www.broadhurst-family.co.uk/lefteye/
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Thu Sep 4, 2008 12:52 am (PDT)

Chris Broadhurst wrote:

correct an image that I have heard about - it was an instant WOW
when I tried it, and it's concept has opened up some new ideas
for me to experiment with. The only weird thing was curving from
a horizontal line!

Yeah, it is a bit different. The idea is that the curve becomes a curve of *differences* rather than a curve of values. (The "linear light" blend mode is basically an "apply differences" mode, in my conception)

Thank you so very much for sharing it with us - I trust you don't
mind, but I've spread your idea around the CurveMeister forum.

Sure, no problem. In that forum, Mike Russell wrote:

I've done some thinking about this, and the procedure seems to be
very similar to using the Lightness channel as a mask, then
curving the L, a, or b channels.

I haven't had a chance to verify this yet.

No, this is incorrect. Using the lightness channel as a mask would merely adjust the amount to which adjustments of curve values show though. That can be useful, but it is much less general/powerful than this technique.

As stated before, this is more like a generalization of curves, channel mixer, applying only based on masks, and blend if sliders, mashed into a single conceptual framework, using a set of curves as the main interface (3, or 9, or possibly more).

A simple example. Imagine I have a navy blue car, and a light red car in my image, on a nondescript background contrasting with both and I want to de-saturate the red car and make it orange, and change the blue car to cyan, and boost its saturation. So I'd take my L*-based adjustment curves (that is, the curves being applied to the linear-light layer w/ L* copied to all channels) and I'd drag the medium-light part of the b* curve up (to add yellow to the red car), the medium-light part of the a* curve down (to desaturate the red car), the medium-dark part of the b* curve up a tiny bit, and the medium-dark part of the a* curve up (to move it cyan-ward). Then I could lighten or darken these cars too, by dragging parts of the L* curve up/down, for instance I could make the red car into more of a peachy color if I wanted, etc.

If I was just curving the L* or a* or b* channels, I could try to alter the saturation of a* or b*, but only based on the values in a* & b* but not on the value in L*. Adding a mask there wouldn't give me fundamentally more power. It would just let me scale back the changes in particular parts of the lightness domain.

I'll try to find some good example images and post screenshots sometime soon, so this makes more sense. It's a bit tricky to explain, but is pretty straight forward to actually *do*.

Cheers,
Jacob Rus
___________________________________________________________________________ 3.

good test images? (was: Re: [colortheory] Re: L* channel duplicated
Posted by: Jacob Rus
Thu Sep 4, 2008 12:52 am (PDT)

Jacob Rus wrote:

I'll try to find some good example images and post screenshots
sometime soon, so this makes more sense. It's a bit tricky to
explain, but is pretty straight forward to actually *do*.

(Just throwing this out. And sorry if I'm bombarding this list with more traffic than I should--tell me if the signal/noise ratio gets too low.)

Actually, are there any good collections of "try to correct these" type images, for testing the mettle of new techniques? (e.g. using Camera Raw, using curves in various spaces, applying one channel to another with the "Calculate" tool, using a channel mixer, &c. It would be pretty useful to compare all the various ways of getting to the same kind of correction goals, both for ease in the journey, for number of steps, and for quality of the finished product.

It could maybe even function something like "Project Euler" for aspiring photo retouchers. (Project Euler is a site with increasingly difficult math problems to be solved by writing computer programs. It lets people improve their skills by working through problems at any level of difficulty.) Each of several pages would have an image (CC-SA or similarly licensed), and a problem/mission statement, and then a "solutions" page would show various users' attempts to correct the image with whatever techniques they prefer, with the best performances (based on simplicity, quality, etc.) sorted to the top, and details of the steps taken.

Anyway, just a hare-brained idea. Feel free to ignore it and respond instead to the main thread, which responses are the ones I am most interested in. :-)

Cheers,
Jacob Rus
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: "Mike Russell"
Thu Sep 4, 2008 3:16 am (PDT)

Jacob,

I want to add my encouragement to keep going with this. For a year or more I've been thinking of ways to combine curves with channel mixing, and what the practical benefits of this would be. I haven't wrapped my mind around your technique yet. It may well be something new, or turn out to be an interesting repackaging of some of Dan's techniques.

Either way, kudos to you for coming up with something that feels genuinely new.

This may well be a missing piece to the puzzle I've been working on. Very exciting, for a curve-centric person such as myself.

BTW you mentioned the desire to scale the vertical curve axis. Would changing the transparency of a curve adjustment layer accomplish this?

Mike Russell - www.curvemeister.com
___________________________________________________________________________ .

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Thu Sep 4, 2008 4:12 am (PDT)

Mike Russell wrote:
I want to add my encouragement to keep going with this.

Thanks. :-)

Let me know if you have any ideas for how to use this with other techniques, or to extend it, or if you wind up with any good example images where it works better than existing tools.

BTW you mentioned the desire to scale the vertical curve axis. Would
changing the transparency of a curve adjustment layer accomplish this?

Oh, check my second email. The way to scale the vertical axis is to add a curves/levels layer on top of our stack, also set to be a clipping mask, which reduces contrast (though setting the transparency would work too).

The best thing with adding a contrast-reducing levels/curves layer is that we can do it to a different degree for the L* channel and for the a* and b* channels. I've decided for my purposes, restricting the adjustments to between about -25 and +25 on the a* and b* axes, and to between 25 and 75 on the L* axis, gives me a pretty decent range, such that the curves I want to make still have a bit of room at the top, but also aren't all completely squished into the middle.

And actually, it's even possible to make this "zooming" curve into something other than a straight line, if you want finer control of near-neutral changes, or similar (I found this less intuitive than keeping things linear in my experimenting).

Cheers,
Jacob Rus
___________________________________________________________________________ .

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Stephen Marsh
Thu Sep 4, 2008 4:17 am (PDT)

Jacob Rus wrote:

* Convert the image to CIELAB (obviously)

* Copy the L* channel (I

* Create a new layer, and paste the L* channel into all three
channels (L*, a*, and b*).

Jacob, I like to try to visualise things in my mind before I first get "hands on".

If I am following you correctly, then we have the data prepared for a channel blend at this point. As the blend is based on the L channel it is very contrasty (extreme) when compared to the subtle colour channels.

* Set the blend mode of this new layer to "Linear Light,

* Add a new curves adjustment layer, and set it as a clipping mask.

We now have a layer/blending mode channel blend at this point.

* Make all three curves--L*, a*, and b*--completely flat, i.e.
Input 0 -> Output 50, In 100 -> Out 50 in the L* channel,
and In -128 -> Out 0, In 127 -> Out 0 in a* and b*. At this
point you should have an image that looks exactly like the
original, because Linear Light blend mode makes no change to
the bottom layer where the blend layer is middle gray.

The effect of the channel blend (tonal values either side of mid gray) is "removed" to a latent mid gray value via some grouped curves.

The (temporary) result at this point in time should be the same as a solid fill layer in mid gray set to LL blending.

* Now here's the fun part: make some points along those curves
and slide them up and down.

Altering the curves now allows areas of the flat gray to become toned at different values than the neutral mid gray value (no longer flat), with the tone being based on the blend of the original L channel data which has been remapped via curves.

The blending mode and curves combined with this channel blend allows one to "dial in" the channel blending effect from "zero" - which is the opposite approach to starting with the blend at full intensity and then lowering the intensity to taste.

Is that correct Jacob?

Sincerely,

Stephen Marsh
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: "Eric Loots"
Thu Sep 4, 2008 8:23 am (PDT)

Hi Jacob,

Kudos for coming up with this approach. A fine example of thinking out of the box.

As far as the scaling of the vertical axis is concerned, the suggestion in Mike's message to reduce the opacity of the curves adjustment layer doesn't work because what it does is kind of like a fading between the applied curves and the default curves.

What does work as an alternative to the approach you explained (using an additional curves adjustment layer. Haven't tried that one yet) is to modify the transparency of the 'linear light mode' layer. For example setting the transparency to 50% makes the curves half as sensitive giving more fine-grained control on the modifications applied in the adjustment layer.

As for this method in general, it will take some time to fully explore and understand what its possibilities are.

Regard -- Eric
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Thu Sep 4, 2008 8:25 am (PDT)

Stephen Marsh wrote:

Jacob, I like to try to visualise things in my mind before I
first get "hands on".

If I am following you correctly,

Yes, your summary is correct. Let me try to clearly lay out the goals here.

What we are looking for is a way to adjust L*, a*, and b* values for particular parts on the input image. Where in this case the amount of the adjustment, and the particular parts adjusted depend on 2 factors: (a) where in the L* domain (or some other channel) a color lies, and (b) how we have set up our curve.

We want to be able to say (for instance), I want every color such that 20 < L* < 40 to have a* increased (more magenta), and every color such that 30 < L* < 50 to have b* decreased (more blue), but I want the effect to be tapered at the edges so that it's not a drastic cut-off.

Okay, so how do we accomplish this? The first way is to use a normal pair of curves layers, one which increases a* everywhere, and then has its "blend if" sliders set to restrict the range to only that particular L* range, and a second which decreases b* everywhere, and similarly restricts the range using blend if sliders. This is great if the kind of difference we want here is constant across the L* range, with linear fall-off between the split ends of the blend if sliders. But if we want something more varied or complex, we need to start making tricky masks, so instead, I got to thinking about what tools we have, and what tools we'd need.

Well, I realized that the linear light blend mode adjusts a source color linearly based on the difference of the target color and middle gray. So if I want to adjust b* up by 2 units everywhere, I want to make a flat color of (50 L*, 0 a*, 2 b*), and then apply that with linear light.

Now then, my question at this point was, okay, how do I make an image which is middle gray most places, but yellowish gray for a specific range of L* values in the source image. Well, of course, we can just apply a curve to the L* channel of the source image which is flat across, but a bit higher in the region we want. Then we can copy that "mask" made of our "curved" L* channel into the b* channel of a new layer, with L* and a* set completely neutral there, and we can set the new layer to linear light. Voila!

And then I thought about this, and realized it was foolish to do this copy/pasting, if I could just apply the curve in the b* channel itself, *after* I'd pasted the L* channel into it. So then I thought, well heck, why don't I just copy the L* channel into all three? Then I could do all kinds of crazy stuff! And the linear light blend will treat each channel independently, so it's like combining 3 of these mask-curve-mix things into one.

Maybe that makes some sense? If not, I can try to clarify any confusing bits.

then we have the data prepared for a channel blend at this
point. As the blend is based on the L channel it is very
contrasty (extreme) when compared to the subtle colour
channels.

Yes.

* Set the blend mode of this new layer to "Linear Light

We now have a layer/blending mode channel blend at this point.

Yes.

The effect of the channel blend (tonal values either side
of mid gray) is "removed" to a latent mid gray value via
some grouped curves.

The (temporary) result at this point in time should be
the same as a solid fill layer in mid gray set to LL blending.

That's right. The adjustment "group" is now just a solid middle gray, which means since it's set to Linear Light blend mode, it should have no effect on the source layer.

Altering the curves now allows areas of the flat gray to
become toned at different values than the neutral mid gray
value (no longer flat), with the tone being based on the
blend of the original L channel data which has been
remapped via curves.

With the "areas" being based on the original L* channel data, yes, and with the adjustment being based on the curve.

The blending mode and curves combined with this channel
blend allows one to "dial in" the channel blending effect
from "zero" - which is the opposite approach to starting
with the blend at full intensity and then lowering the
intensity to taste.

Yes, exactly.

It allows us to adjust the difference from neutral we want, at each level of the "adjustment channel". Actually, one thing I could use here, to make this clearer, is some well-defined terminology for the parts I'm talking about, and a good name for the technique as a whole. Ideas?

Cheers,
Jacob Rus
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: "klaus_nordby"=
Thu Sep 4, 2008 12:45 pm (PDT)

=Jacob Rus=wrote:

Actually, one
thing I could use here, to make this clearer, is some
well-defined terminology for the parts I'm talking about,
and a good name for the technique as a whole. Ideas?

I've already, in my own testing of this radical up-down-curve approach, given it the perfect name: "Jacob's Ladder". :-)

Klaus Nordby
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Sat Sep 6, 2008 2:19 pm (PDT)

Okay, I'm happy to use that until I hear something better. :-)

I made a "Jacob's Ladder" [action][1], which creates 3 new linear light layers, 3 "adjuster" curves layers, and 3 "zoom" curves layers which scale the effect of the adjuster layers so we can use more of the vertical space of the curves dialog.

So, the parts to actually play with here are the "adjuster" layers (colored red). In general, the other layers should be left alone.

Also, I took some screenshots while playing with the "Jacob's ladder" curves of a cc-by licensed flickr image, and strung them together into an [animated GIF][2].

Cheers,
Jacob Rus

[1]: <http: //www.hcs.harvard.edu/~jrus/colortheory/Jacob%27s-Ladder.atn>
[2]: <http: //www.hcs.harvard.edu/~jrus/colortheory/jacob%27s_ladder.gif>
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Sun Sep 7, 2008 6:41 am (PDT)

Okay, I've extended that action set to also include actions called "Jacob's RGB Ladder", "Jacob's CMY Ladder". These leave the image in CIELAB mode, but make "adjusters" which are based on the RGB, respectively CMY, channels. Also, I added a "Jacob's Generic Ladder" action which brings up the apply image dialog to allow a choice of source channel/mask to use.

The only issue is that adding a "convert to profile" step with a custom CMYK separation (No-black GCR with SWOP uncoated ink) balloons the file size by 600+ kilobytes. Which is odd, because the instructions for that space take up only a few text fields. Oh well.

Hopefully those are useful to someone. I'm still planning to make a more detailed tutorial, but I'm not sure when I'll have time: school starts in about a week, and I'm flying from CA to Boston on Tuesday night. :-)

Cheers,
Jacob Rus
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: "Ric Cohn"
Mon Sep 8, 2008 9:03 am (PDT)

Jacob,

Thanks for posting all this information. I don't get much of it yet, and I can't say whether there are other ways to do the same thing, but it sure is interesting.

Two difficulties I do see which you might be able to address:

1. the a and b can be very noisy and I'm not sure how amenable this method is to blurring of the channels.

2. While working on the Linear Light curves the info pallet info is wildly inaccurate. As far as I can see, this rather limits its control and accuracy to judging visually or to iterative adjustments.

Keep up the interesting work.

Regards,

Ric Cohn
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Jacob Rus
Mon Sep 8, 2008 10:10 am (PDT)

Ric Cohn wrote

Two difficulties I do see which you might be able to address:

1. the a and b can be very noisy and I'm not sure how amenable this
method is to blurring of the channels.

Sure, you can blur the channels used for modification, or you can blur the channels after you're done with your edits, etc. etc. The noise reduction required here is comparable to other methods which base lightness on color info. I admit though, this technique makes such mixing easier, and output noise deserves some consideration.

Especially in big areas of similar color (the sky, etc.), just blurring the noise in the b channel works quite well.

2. While working on the Linear Light curves the info pallet info is
wildly inaccurate. As far as I can see, this rather limits its
control and accuracy to judging visually or to iterative adjustments.

Basically the problem here is that when working in a clipping mask, or in a layer set (folder) which is set to some blend mode other than Pass Through, Photoshop decides to display the values in the info palette for the current clipping group or layer set, rather than for the entire image, but only while you are currently working in a modal adjustment view.

As far as I'm concerned, this is a bug in Photoshop (I can imagine someone wanting to see this info in the info palette, but the percentage of the time when this would be useful is vanishingly infrequent compared to the percentage when showing the info for the composite image is preferable). If anyone knows a workaround, please share.

If you dismiss the modal dialog, you'll see "correct" numbers in your info palette.

-Jacob
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: "Ric Cohn"
Mon Sep 8, 2008 2:56 pm (PDT)

On Sep 6, 2008, at 9:00 AM, Jacob Rus wrote:

[1]: <http: //www.hcs.harvard.edu/~jrus/colortheory/Jacob%27s-Ladder.atn>

Jacob,

I believe there's an error in the RGB and CMY scripts. When they call for a duplicate file it is named "Untitled-1" by default (not sure if the numbers will increase if the script is applied to more than one image). The script is looking for a document named "jacob's ladder temp". When I use this name for the duplicate document the script appears to work.

Regards,

Ric Cohn
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: "jacobolus"
Wed Sep 10, 2008 9:41 am (PDT)

Ric Cohn wrote:

I believe there's an error in the RGB and CMY scripts. When they call
for a duplicate file it is named "Untitled-1" by default (not sure
if the numbers will increase if the script is applied to more than
one image). The script is looking for a document named "jacob's
ladder temp". When I use this name for the duplicate document the
script appears to work.

I just tried my action with Photoshop CS3, and apparently that's the issue: CS3 doesn't allow making a step in an action which just creates a new image of the same size and color space as the clipboard (and if you have such a step in an action created with CS2 it pops up a dialog instead of just making the new image). I am not sure why this is, but I'd consider it a regression.

Maybe someone with more action-making experience has advice here?

-Jacob
___________________________________________________________________________

How Does Selective Color Work?
Posted by: Howard Smith
Wed Sep 10, 2008 9:06 pm (PDT)

To the best of my knowledge, this question has not been posted on the Forum.

How does Selective Color work? It seems to target specific colors in a way that Curves cannot. Can someone tell me how it does that?

Howard Smith
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: Michael Jahn
Wed Sep 10, 2008 11:48 pm (PDT)

Hi Howard,

I use this (Selective Color) when I have to change a specific color in an image (like a blue dress to green) without creating a mask.

http: //www.newtutorials.com/selective-color-adjustment.htm

or -- are you asking what approach (formula / algorithms) are used ?

in a word -- LUTs (Look Up Tables) - (okay, that was three words...)

in many words;

from United States Patent 5506946 - Selective color correction;

==========================

A method and apparatus for modifying colors in an image displayed on a display device by a computer controlled display system. The method includes selecting at least one color being representative of at least one pixel of the image, where the image includes a plurality of pixels on the display device; determining the colorimetric values for the at least one color; selecting a second color and determining the colorimetric values for the second color; and modifying the colorimetric values of a plurality of pixels in said image such that for any given pixel of the plurality of pixels having colorimetric values being matched to the colorimetric values of the at least one color, then the colorimetric values of the given pixel are modified to match the colorimetric values of the second color. The apparatus of the invention includes a display device; a cursor positioning device; a plurality of smoothed, look-up tables to provide a selection of a region of a color space, which selection is representative of a portion of said image near the at least one color; memory for storing data representative of the region and a second color; and a data processing unit. The method and apparatus allow a selection and automatic modification of a group of colors while maintaining the relationship among the colors so as to appear relatively the same with respect to texture and tonality.

==========================

Wow, yeah, that made little sence to me - even afer like 9 or 10 times...

Like anything and everything -- it is an A to B conversion. All pixel values - or combination of pixel values are somewhere between 0.000 and 1. You want to make a change in that value, no matter if you use some curve bending interface, some levels like interface or some slider bar interface, all you are doing is point that original pixel value to another pixel value and saying 'make this like that'.

you can do these sorts of things in JavaScript (without Photoshop) as it is all A to B.

it is just math. complex math. okay, complex math by color scientists.

with sugar on it.

Michael Jahn
PDF Conversion Specialist (PDF files often have images in them)
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: Stephen Marsh
Thu Sep 11, 2008 2:21 am (PDT)

Howard Smith wrote:

How does Selective Color work? It seems to target specific colors
in a way that Curves cannot. Can someone tell me how it does that?

Howard, those that know are probably under an NDA. I like to know these "under the hood" things too, although often it is a trade secret. One can try to reverse engineer things or come up with test files to help understand various operations.

That being said, independently of the document colour mode, it appears to work in Lab space behind the scene (on the fly?), in order to target various hues.

Using complex ChOps, it is also possible to select various hues, using the R, G, B channel data for the blends - although this is only for targeting the hues in question, not changing their values.

Just as Selective Colour can isolate various hues, so too can other commands in Photoshop (Hue/Saturation, Color Range, black and white command) - which as you say is different from a global curve move.

So I think that there are two parts to your question; one is on isolating certain colours and then, secondly, editing the isolated colours.

Sincerely,

Stephen Marsh
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: "Lee Clawson"
Thu Sep 11, 2008 10:01 am (PDT)

Howard,

In addition to what Stephen Marsh wrote I'd like to add another thing to watch out for.

Selective color looks at the color by the numbers. This is to say that even if the color looks like a simple blue "selective color" will look at all the colors in your image that have some form of blue and affect them accordingly. In addition if you have any gray component in your blue its possible for Selective color to see that as a neutral and affect those too.

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: "John Bongiovanni"
Thu Sep 11, 2008 4:04 pm (PDT)

Stephen, could you elaborate on this or give and example (or two)?

John Bongiovanni
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: Stephen Marsh
Thu Sep 11, 2008 8:56 pm (PDT)

I wrote:

That being said, independently of the document colour mode, it
appears to work in Lab space behind the scene...

Apologies Howard/List, I was thinking of Select/Color Range and not Image Adjustments/Selective Color.

Howard, are you seeking info on using the controls, or how the CMYK based interface works in CMYK and RGB modes (each having different channel structure and editing response).

Sincerely,

Stephen Marsh
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: Stephen Marsh
Fri Sep 12, 2008 10:27 pm (PDT)

John Bongiovanni replied:

Stephen, could you elaborate on this or give and example (or two)?

John, let me reply to this by asking another list member some questions.

On Tue Sep 2, 2008 - Jacob Rus made the following post:

L* channel duplicated across a* and b* for finer color control
http: //tech.groups.yahoo.com/group/colortheory/message/20162

My questions are for Jacob Rus.

Jacob, depending on your answer to these questions, it may perhaps be better for you to answer John's question!

1) Jacob, the Lab mode based LL channel blending curve method that you posted on September 2 - is this your own original work?

2) If not original, is this basic technique adapted from another source, with some unique additions of your own over the original concept?

3) Have you posted this method previously to other lists, or was the post noted above your first public announcement of this method?

4) If you post on other online websites/forums/boards/maillists etc., do you go by the name of Jacob Rus or do you use a different 'screen name'? (I am not referring to Mike Russell's Curvemeister forum)

[P.S. Although I am a list modertor, these questions are *not* list moderation questions]

Sincerely,

Stephen Marsh
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: Jacob Rus
Sat Sep 13, 2008 8:21 am (PDT)

Stephen Marsh wrote:
 
Jacob, depending on your answer to these questions, it may perhaps be
better for you to answer John's question!

1) Jacob, the Lab mode based LL channel blending curve method that you
posted on September 2 - is this your own original work?

2) If not original, is this basic technique adapted from another
source, with some unique additions of your own over the original concept?

Yes, indeed, I had never seen anything like this discussed before, as I said in my original post.

3) Have you posted this method previously to other lists, or was the
post noted above your first public announcement of this method?

Nope, it was the first announcement.

4) If you post on other online websites/forums/boards/maillists etc.,
do you go by the name of Jacob Rus or do you use a different 'screen
name'? (I am not referring to Mike Russell's Curvemeister forum)

I go by the handle "jacobolus" on IRC, various instant message services, Wikipedia, various online fora (such as the ars technica fora, or the few posts I have made on that curvemeister forum), &c. (and you'll note my emails to this list are sent from jacobolus.ml@gmail)

Also, I put up a page with a few example corrections:
<http: //www.hcs.harvard.edu/~jrus/colortheory/jl.html>

Cheers,
Jacob Rus
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: "Rick Gordon"
Sat Sep 13, 2008 7:35 pm (PDT)

Taking nothing away from Jacob's intriguing procedure, which takes the concept in some different directions, I might mention I posted this article to the list on July 12, under the title of "Another Local Contrast Technique". There is some similarity in the approach:
___________________________________________________________________________

To: Color Theory List
From: Rick Gordon
Date: Sat, 12 Jul 2008 20:10:02 -0700

Subject: [colortheory] Another Local Contrast Technique

I've been exploring another local contrast technique a bit different from others that I have heard about, so I thought I'd share it and see if it's being used by or of use to others. I have set this up as an action.

1) Start with either

A) a duplicate (either color or desaturated) of the image layer OR

B) a copy of a channel, OR

C) an idealized built channel, created by blending or other approaches. (More on this below.)

2) Set its blending mode to a strong overlay blending mode; I like the effect of Vivid Light, in particular; and especially if the layer for this work is desaturated or a copy of a single channel. The choice of blending mode can have an effect on final color (but not the way you might think). I may be useful to supply a copy of the original file above the entire stack, set to Color mode, which can be adjusted for opacity at the end.

3) Set up a Curves layer, set into a clipping group with the previous layer, in Normal blending mode, set to a default (no change) value. This will be useful in the layer stack for manipulation at a later point.

4) Set up a Solid Color adjustment layer, set into a clipping group with the previous layer, with the color set to a 50% gray (128,128,128). If you prefer, a standard layer filled with 50% gray would have the same effect.

5) Back off it's opacity slightly, perhaps to about 95%, to taste.

6) Adjust the opacities of Vivid Light layer to taste.

7) The Curves layer set up in step #3 can be used to adjust shadow points and make slight adjustments in other tonal ranges. You will find that large moves are very subtle. Even though it is set to a Normal blending mode, you will find that that operations on individual channels will not affect color balance. Unusual direction-changing curves can be useful here. Experiment. It will be useful to understand that a horizontal straight-line curve (128/0, 128/255) will essentially cause the Vivid Light layer to act as if it's been turned off as far as luminosity changes go; there will be a very slight saturation effect, keyed to the opacity level of the 50% gray layer above it.

The effects of this action are surprisingly subtle and controllable, which may be surprising if you have dismissed the use of Vivid Light or Linear Light blending modes as only useful for over-the-top hack effects.

I have some more ideas that relate to this approach, but within the context of a high-pass sharpening procedure; but I'll save that for another post.

___________________________________________________

RICK GORDON
EMERALD VALLEY GRAPHICS AND CONSULTING
___________________________________________________
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: Jacob Rus
Sun Sep 14, 2008 4:22 am (PDT)

Rick Gordon wrote:

Taking nothing away from Jacob's intriguing procedure,
which takes the concept in some different directions, I
might mention I posted this article to the list on July 12,
under the title of "Another Local Contrast Technique".
There is some similarity in the approach:

Yes indeed. Neat! I wouldn't claim that the idea as I have expounded is particularly earth-shattering: it is basically a way to do straight-forward manipulations of one channel based on the contents of another channel, using a familiar interface.

(What is more surprising is that Photoshop doesn't have more tools for this sort of thing built in; many of the color tools Photoshop does provide have cryptic simplified sliders whose operation is unclear from description alone, and must be painstakingly deduced from extensive practice. I prefer well-understood tools whose abstractions allow me to express the often-complex transformations I can imagine in my head with minimal fuss; flexibility and generality are more important than basic learning curve, if the effects are predictable and comprehensible).

4) Set up a Solid Color adjustment layer, set into a
clipping group with the previous layer, with the color
set to a 50% gray (128,128,128). If you prefer, a
standard layer filled with 50% gray would have the
same effect.

5) Back off it's [sic] opacity slightly, perhaps to about
95%, to taste.

Ah, interesting: another way to handle effectively "zooming" the curve in the curves layer, besides adding transparency to the underlying layer, or (as I have done) adding a flattening curves layer on top of the main curves adjustment layer (though in 16-bit mode, I like my zoom approach better, for two reasons: (a) I can construct the two curves such that 0 in the output of the main adjuster results in no change to the underlying layer, whereas zooming with transparency or a mostly-opaque gray layer on top leaves a setting of 0 on that main curve actually corresponding to a slight positive change; and (b) I can adjust the amount to zoom differently for different channels, for instance zooming the color channels more than the lightness channel.

6) Adjust the opacities of Vivid Light layer to taste.

This step is redundant with the previous two steps (the two approaches have identical effect)

7) The Curves layer set up in step #3 can be used to
adjust shadow points and make slight adjustments in
other tonal ranges. You will find that large moves are
very subtle.

This is only true to the extent that the base layer of the clipping group is transparent or the gray layer on top is opaque. The goal should be to make the useful range of adjustment correspond to the height of the curve.

Even though it is set to a Normal
blending mode, you will find that that operations
on individual channels will not affect color balance.

I'm not sure I understand this. Operations on individual color channels in RGB certainly affect color balance.
Unusual direction-changing curves can be useful
here. Experiment. It will be useful to understand
that a horizontal straight-line curve (128/0, 128/255)
will essentially cause the Vivid Light layer to act
as if it's been turned off as far as luminosity
changes go [...]

Yep. And this is why I'd recommend flat as a starting point.

I think the main difference between what you've described and what I have described is in overall approach to the problem, rather than in possibilities enabled. I had a very specific goal of modifying one channel based on another, using a curves tool as an interface, and realized I could in fact combine three such modifications into a single clipping group, or 9 into a set of 3 clipping groups, and thereby create a substantial new general-purpose tool. It seems like your goals were more general, or perhaps less precisely formulated (this is not intended as criticism).

[...] if you have dismissed the use of Vivid Light
or Linear Light blending modes as only useful for
over-the-top hack effects.

Yes, Linear Light can be useful for other purposes too, I'm sure; linearity of modification is a useful property when using curves adjustments, because it is extremely predictable.

I think the reason many users dismiss some of the blend modes is that most users just try them without understanding how they work or considering how they might be made to achieve desired effects, and try them on normal images, rather than layers specifically constructed to be used with them. Here's a good picture of the various modes in action: <http://dunnbypaul.net/blends/>

I have some more ideas that relate to this approach,
but within the context of a high-pass sharpening
procedure; but I'll save that for another post.

What would you do for sharpening? I must admit I have no particular sharpening expertise. I do some combination of increasing local contrast with large-radius sharpening, plus close sharpening of detail before final printing, with most images, usually in the L* channel, but there are likely refinements which would allow more perceived sharpness without other artifacts (halos &c.). I imagine there has been significant discussion of that topic on this list, previously.

Cheers,
Jacob Rus
___________________________________________________________________________
.
Re: How Does Selective Color Work?
Posted by: Howard Smith
Sun Sep 14, 2008 2:15 pm (PDT)

First of all, Stephen, I want to thank you, Michael Jahn, and Lee Clawson for your thoughtful answers to my question. It seems that Selective Color is just a bit more complicated than expected. At least it's a comfort knowing that there wasn't something obvious there that I had overlooked.

My reason for asking the question was to learn more about the workings of Photoshop. A countless number of posts on this Forum have been invaluable in adding to my store of knowledge. Sometimes just a single sentence has led not only to a greater understanding of Photoshop, but to exciting new approaches to color correction. My hope was that Selective Color might contain similar bits of information that could be incorporated into future work. In the beginning, some years ago it seems, I firmly believed that Dan was saying that Curves is the only tool one needs for professional color correction. Big mistake on my part, though my struggles with Curves did prove fruitful in the long ruin. This is one of those tools with so much hidden potential that it's unlikely that any of us will ever discover all of it. Slowly but surely other tools are being added to my collection, Selective Color being another of those with much more potential than one might suspect. As a recent example, I've long had problems with Cyan contamination of yellows in my images, leading to green fluorescence with Epson pigments. Reduction of Cyan invariably led to greater expression of its complement, Red. A red cast wasn't much better than a green cast. But with Selective Color it's possible to reduce Cyan and keep red under control at the same time, resulting in an image with the same warm colors but with much less of the contaminating Cyan.

Thanks to all of those who respond to my sometimes simplistic questions. The help has been invaluable.

Sincerely,

Howard Smith
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: J Walton
Sun Sep 14, 2008 3:11 pm (PDT)

On Sun, Sep 14, 2008 at 8:29 AM, Howard Smith wrote:

First of all, Stephen, I want to thank you, Michael Jahn, and Lee Clawson
for your thoughtful answers to my question. It seems that Selective Color
is just a bit more complicated than expected.

Not really - there's not much to it. Selective color is akin to some classic color correction controls typically found on high-end drum scanners. Since you could not make selections on drum scanners and overall curves do not solve every problem - particularly with saturated colors - selective color was a necessary part of scanning.

Here's how it works in CMYK:

Where Photoshop sees a color as being mostly cyan, you are able to adjust all 4 colors (C M Y K) in that range by selecting "Cyans" from the pulldown menu. Because nothing in reality is purely cyan, these moves are never really "full strength" - they are tempered by the fact that the colors you are trying to adjust are partly cyan and partly something else.

There are two different basic settings in Selective color (I forget what they are called, they are radio buttons at the bottom of the dialog box). One is Absolute, and I think the other is Relative. Relative is more like what scanners used to do, and is the more elegant of the two. Relative will work more like a midtone curve would - if there's something there it will move otherwise it won't. So you can't add yellows to a pure cyan using Relative because they aren't there. You can, however, make a nice move to magentas in a blue because they are approximately in midtone. Absolute works more like working from the highlight or shadow end of curves, and is considerably less elegant. So you can add cyan to a pure yellow even if there was no cyan to work with.

Even though Selective Color is by definition a CMYK tool, I find it more powerful for adjusting saturated colors when using it in RGB. You can run some quick comparisons to see the effect, but they tend to be much larger and more dramatic in RGB. And because I "grew up" on CMYK I think those numbers anyway, so it still makes sense to me.

Here's some good uses for Selective Color:

Blacks - in CMYK you can have a nice amount of under color control, though perhaps not as much as if you used Channel Mixer. Removing CMY from Blacks is a classic dmax fix, and is nice because you don't need a mask and it can sit on top of all other adjustment layers.

Flesh - when skin is going red in one part and yellow in another, I've had good (and especially QUICK) results in adding magenta to yellows and removing it from reds. Obviously you'll want to look at cyans in reds and yellows in yellows as well.

Whites - This can be a quick way to get shape in lighter areas, even a flesh tone. In RGB, remove black from whites. Since there is no black in RGB, the black sliders work with overall lightness and darkness.

Casts in saturated colors - this is the real point of selective color. After fixing overall balance problems, if you still have some issues with saturated colors it is easy to target them using selective color.

Here's some lame stuff about Selective Color:

You can't adjust the range of colors that are being affected, as you could with Hue/Saturation

Neutrals is frustrating because it basically affects EVERYTHING. It would be awesome to actually be able to affect things that are near-neutral. Hue/Saturation in particular should have this capability.

Remember that Selective color is, at best, a 2nd rate color correction tool because it has limited control over the channels. But it's just another arrow in the quiver of a skilled retoucher, so it's a good thing to be familiar with.

Regards,

J Walton
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: "Mike Russell"
Sun Sep 14, 2008 5:16 pm (PDT)

I've been meaning to do this since Howard first posed the question. So this afternoon I played around a bit with this command to see whether, as I initially suspected, Selective Color uses Lab under the hood. Turns out it does not.

As Stephen mentions, another command - Select>Color Range - works in Lab space. This can be verified by creating an Lab image with gradients in the a and b channels that are rotated 90 degrees to one another, and using the command. In Lab mode, the selected colors show up as perfect squares, corresponding to overlapping ranges of a and b values, with the size of the square controlled by the tool's fuzziness setting. Converting to RGB mode gives a similar result. Due to gamut clipping, the squares are only well-defined for a narrower range of a and b values.

Selective color is a different animal. There are no nice squares if you add black to red, for example. Rather, the sections are more or less pie-shaped, pointing toward a hue angle based calculation. Using an RGB color wheel image, selective color affects an equal range of color angle for each of the components RGBCMY. This is characteristic of an HSB based calculation, and not an Lab one. If you turn the Select>Color Range command loose on the RGB color wheel, it uses a larger angle of green than magenta, as would be expected in Lab space.

Mike Russell - www.curvemeister.com
___________________________________________________________________________

Re: L* channel duplicated across a* and b* for finer color control,
Posted by: Stephen Marsh
Mon Sep 15, 2008 1:24 am (PDT)

Jacob I had an (incorrect) feeling that you were somebody else that goes by a different name on a different list!

Thank you Jacob, for confirming that the method was to the best of your knowledge "original" to yourself and that you did not adapt it from somewhere else and also for confirming that you are not the same person that I was thinking of from another site.

Guess what I found when doing a search for a different topic over the weekend?

It seems like this is a case of two researchers coming up with similar methods, at different points in time!

______

http://www.ozoneasylum.com - Photoshop Forum
Post #28789 Lab Bauble
01-03-2007 08:02 by warjournal

- Start with some random photograph.
- Go into Lab mode.
- Copy photo to another layer.
- On the copy, to go Channels palette and C&P L into a and b

That last step is the magic part. Now just gotta finish setting up the rig.

- Change the blending mode to Linear Light so we can add/subtract values.
- Blending Options and turn off L channel.
- Clip a Curves Adjustment Layer to this copy.
- In the Curves, 'zero out' the a and b curves.

Once it's all set, get busy with the tweakage.

You can now very quickly and easily adjust colour balance based on L without masking. That is, based on any L value(s) with perfect blending."

______

Sincerely

Stephen Marsh
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: "Lee Clawson"
Mon Sep 15, 2008 1:14 pm (PDT)

Howard,

One way to see what's happening for yourself is to open a Macbeth or IT-8 color chart. Start with Selective color looking at "Reds", make a 15% or more edit and use the eyedropper/info palette to watch what happens with all the other colors.

Lee Clawson
2/\V/\7 Studio
___________________________________________________________________________

Re: How Does Selective Color Work?
Posted by: Jacob Rus
Mon Sep 15, 2008 8:36 pm (PDT)

As for the question of selective color, this seems to be a reasonable description:

<http: //www.photoshoptechniques.com/forum/showthread.php?p= 202922#post202922>

-Jacob Rus