Color-previews of the "A" and "B" channels

[mikemeister_admin]

Greetings & Happy New Year from Oslo, Norway, ye beta-folks! 

I privately sent this to Mr. CurveMeister, and he encouraged me to post it here.  I've made a LAB channel preview which color-previews the "A" and "B" channels, as shown in this attached file.  I find this somewhat educational in the mysterious ways of LAB.  I've asked Mike to implement something like this inside Curvemeister, somewhere, somehow -- and hope others here would like to see it, too.

Klaus Nordby

[mikemeister_admin]

OK, I tweaked and improved my colorizing procedure a bit and applied it to another test image.

K

[derekfountain]

I privately sent this to Mr. CurveMeister, and he encouraged me to post it here.  I've made a LAB channel preview which color-previews the "A" and "B" channels, as shown in this attached file.  I find this somewhat educational in the mysterious ways of LAB.  I've asked Mike to implement something like this inside Curvemeister, somewhere, somehow -- and hope others here would like to see it, too.

I like this. It's a lot more expressive than the feeble grey contrast normally used to show an LAB image's a and b channels. Then again, how useful is it? I never look at the normal a and b channel images unless I think they might make a mask. The information they contain doesn't seem too useful to me.

One immediately obvious use would be finding neutral areas, or more specifically non-neutral areas. Suppose you have an image containing a road or building that you would expect to be neutral. My normal procedure is to run the mouse over it and watch the hue clock. This type of view Klaus is proposing would immediately show a colour tinge where there's a cast.

It might also help with direct adjustments of colour saturation. I was recently working on an image that contained a lot of grass which I wanted to reduce the saturation of. I've found that surprisingly hard in LAB mode, and often struggle to find the curve adjustments that reduce yellow and green in the appropriate range. I can see that this view would give me more direct feedback as I try to make that adjustment (although it's a lot easier with PS hue/sat adjustment anyway).

So I think I like the idea! But if there were to be a CM implementation I think being able to see the 3 channel images side by side, as opposed to individually as is currently the case, would be essential. I'm not sure how that would fit into the UI.

[mikemeister_admin]

Glad to have you on my team, derekfountain. :-)

It also occurred to me that all the middle-grays could be normalized to white, like here. That's partly helpful --partly not, as the light yellow areas then become nearly imperceptible. But I offer this in the spirit of exploring these newfangled colorized ABs (which I'm gonna (TM)(C)(R), etc. :-)).

K

[derekfountain]

It also occurred to me that all the middle-grays could be normalized to white, like here. That's partly helpful --partly not, as the light yellow areas then become nearly imperceptible. But I offer this in the spirit of exploring these newfangled colorized ABs (which I'm gonna (TM)(C)(R), etc. :-)).

No, go back to grey. Loosing the faint yellows is far too big a loss!

[mikemeister_admin]

OK, so I had another idea in this area: to split the "A" and "B" into "A1" and "A2" and "B1" and "B2".  Is this a world-historical first, or what?!  Irrespective of seniority claims, it's also an illuminating experiment -- because it shows how *little* blue there actually is in the "B" here, as "B2" shows. 

Did I screw up my algorithm perhaps?  No, I don't think so, I have this down to a science now -- for what we are seeing here is the purely *perceptual* phenomenon of simultaneous color contrast.  All the yellow in "B" pushes our perception in the direction of its complimentary color in all the areas which are really only pure gray -- namely blue.  The "B1" contains only yellow and gray -- and yet is seems rather bluish.  The same principle holds for the "A", of course, it's just not noticeable here, since the magenta and green are fairly balanced in this particular example, hence one hue doesn't overpower the other hue like the yellow does in the "B".

Fascinating, huh?  Mike, when are you gonna chime in? You're The Godfather Jr., you know -- and I need your blessing in this little family matter. :-)

[ggroess]

Ok,
Trying to get my mind around this one....What an interesting view of LAB...
Be patient...

The way I see this is that the values of the color shown are the L, A and B each as it is shown in the separate channel. To me the basic problem is seeing the effect of the L channel on the image.  

For instance if you look at the color checker square...the second row from the bottom second from the left.  Dark Green...
I read it as a little green, a little yellow, I still have to keep the L value in mind as I adjust the image in my mind....

I'm not sure where this leads me from an adjustment standpoint.  
My Questions are as follows:

Would I see the composite as I change the A or B value?
Would the Curves be displayed over this image in stead of the histogram?  
Mike, could that be a choice? I might prefer this to the histogram and I might give me more visual data as to what my changes are doing.  
What tonal value would you use as the background for the A or B mask and would that contribute tonal values from the L channel?

When do I get to see a version of this in a beta release to see if there is any more I can do with it??

Greg

[mikemeister_admin]

Klaus, this is a great illustration! It is quite satisfying to see the channels in color. My work in the individual channels, however, has been limited to making masks, and I am generally thinking in terms of black and white in that case. I tried adjusting a mask in CM while neutralizing the other two channels, or with one color channel and the lightness channel, and it did seem easier to work with the edges of the mask, which resulted in a mask in which I didn't need much (if any) blur applied.

I did notice something as I was fooling around—I can create a very similar color channel in CM by leaving one curve alone, and drawing a horizontal curve at "0" with the other two channels. When I bring these into Photoshop along with yours, the info palette reads "0" for the excluded channel in the CM version, while in your demo, there is quite a bit of info in the excluded channel.

What's going on with that? Regardless, I would like someone to think of a really good reason to include this in CM. I nominate Mike, though he looks a little distracted today.

:)

[mikemeister_admin]

Greg, I have no idea, at present, how this could or should or might be used inside CM. I have NOT offered it here even as a feature-request idea and I'm NOT suggesting that every program should start colorizing the AB channels.. My sole point, so far, has been to explore this idea of colorizing the AB channels -- to make those well-known weird grayscale A & B guys a bit more comprehensible to us lowly, non-geeky artists and photographers. It's up to Mike to ignore this or implement this as he chooses. I just think there's potential here for making the LAB color space a little bit less obtuse and weird -- and if some of it could end up in CM, that would be thrilling.

[mikemeister_admin]

gwilson, thanks for your remarks! I'll look into what you're talking about. As you see, I have opted to use the more standard Magenta and Green RGB colors for the A colors, instead of those you display -- which may or may not be more "correct." But what exactly are the "correct" AB primaries -- given that they are imaginary colors, outside of human-vision gamut? If I understand LAB correctly, that is, which is hardly a sure thing!  But perhaps the ever-absent Mike will chime in here one fine day?

(PS: My apologies if this should be posted twice here now, as I had some connection troubles)