Color Adjustments in LAB, Part II
LAB by t he N umb er s:
In this tutorial's image, we're going to need to analyze some
numbers to get a better idea of what's going on before we can fix it. So let's run through how
the numbers work in LAB.
Note: If you're not familiar with how LAB numbers correspond to colors and brightness, please
spend some time with this first section. A lot of information is gone over very quickly here and
the more comfortable you are with it, the better you will be in making LAB adjustments.
Lightness: The range is 0-100 with zero being black, 100 being white, and 50 being a middle
gray (well, not exactly middle gray, but never mind all that right now).
The A channel: This runs diagonally on the color wheel from magenta to green. The numbers
range from -128 to 127 (to make a total of 256, including zero). The minus colors in LAB are
always the cool colors (green in the A channel, blue in the B — see the color wheel) and the
plus numbers are always the warm colors (magenta in A, yellow in B). A color of zero is,
well...not a color at all. An A channel value of -10 is slightly to the green side and parentheses
will represent minus as in (10A). A radioactive magenta would be +100 shown as simply 100A.
The B channel: Yellow values run from 1B to 127B, while the blues range from (1B) to (128B).
Remember, the parentheses mean minus.
The A and B channels: Zero in channels A and B would be no color at all. So, 0L, 0A, 0B would
be black, etc. A value of 70A, 70B would be a mix of magenta and yellow in equal proportions
producing a reddish hue (but see The Disclaimer, below). A cyan would be something like
(30A), (30B). What would a hue of (3A), (25B) be? The minus B channel pulls the color strongly
into the blue, while the minus A channel pulls this blue slightly to the green side. Thus, the
color would be a slightly cyanish blue.
One final test. What would 84L, 15A, (21B) look like? It would be a very pale purplish color. The
(21B) pulls into the blue, but the 15A drags this slightly up towards magenta producing the
purple color. The 84L is very bright and will produce the pale tinge.
The Disclaimer: Predicting colors in LAB based solely on the A and B channel values can go
horribly wrong very quickly. In the 70A, 70B example, above, if your L channel is very high, say
90 or so, the color will not be red at all — it will be a pale orange! If you think green will be
produced with something like (-128A), 0B, you'd be dead wrong — you'll actually end up with a
teal! To get the green green that we all know (a Prell shampoo green, or a pool table green),
you need to get numbers like 60L, (60A), 60B! I urge you to bring up your Color Picker, click the
L radio button, and start punching in numbers to get a better feel for what's going on here.
St ep One:
Before we start looking at numbers in our
image, bring up the Threshold dialog (Image > Adjustments > Threshold) and find the darkest
and lightest areas.
Now bring up the Info Palette and check out the numbers for these places by simply running
your cursor over the image and reading the numbers in the Info Palette. Circled in green is the
darkest area and this was coming in at 1L, 0A, 0B (circled in red). The Lightness of 1 is too dark
— it's basically completely black with no detail. I'd prefer this to be somewhere around 6 or so.
The lightest area was near the very top, to the right of center. This was giving values of 91L —
not light enough. Here, I want something around 96.
The numbers tell me pretty much what I see — the entire image needs to be lightened up.
St ep Two :
Moving over the rest of the image was giving numbers
that were basically a disaster. There were no colors whatsoever in the positive values — no
magentas, no yellows. Everything was a blend of blues and greens. The snow, circled in red,
was giving strong negative A and negative B.
The road, interestingly enough, was giving something like 60L, (16A), 2B — a grayish green! You
never would have guessed that by looking at the image!
The analysis of these numbers tells me pretty much why this image is so flat — not only is the
brightness range too limited, but virtually every single color falls in the blue/cyan/green
range, with the vast majority of them being a kind of grayish cyan.
It's pretty obvious now that in our A and B channels work, we need to pull away from the blue
and green and pull in slight amounts of magenta and yellow.
St ep Thr e e:
As you can see from the A channel, this is exactly
what I did. The bottom point was brought to the right (toward magenta) by 3 intervals. The top
point was taken to the left (toward green) by about 2.8 intervals.
Our line does not now cross through the center of the grid. It is slightly to the right of the
center point, and just what I want — away from the green and toward the magenta.
St ep Fo ur :
I did basically the same thing in the B channel. The
bottom point was pulled over by only 2 intervals (toward yellow), but the top point was moved
over only 0.8 intervals (back toward blue — but not nearly as much as the yellow came over).
Our line now passes significantly to the right of the center point — more than in the A channel.
I needed to do this because of the strong blue/cyan cast to the picture.
St ep Fi v e:
Now on to the Lightness channel.
Beyond the fact that I knew we needed strong lightening here, the adjustments were
completely subjective (as with all my other adjustments).
I pulled the bottom point straight over by almost 3 intervals and left the other point alone!
Fi nal I mag e:
And here's the result. The blue cast is almost completely
gone. The snow in the original image was pulling strongly negative A and B values — minus 10
to minus 20. In the final version, the A and B were hitting slightly negative: -2 to -5.
Other parts of the image were now showing positive A and B values making the evergreen
trees, at least, look much better.
The image still needs work, though! The brightness range got a bit blown out. The lightest
parts now hit 100, too much, although the darkest moved up a bit from the original 1L.
Working with Shadow/Highlight in the Lightness channel will bring out the clouds to a much
greater degree, and then sharpening (also in the Lightness channel), will make even further
I think you'll agree, though, the difference is nothing short of amazing. And the best part? The
total amount of time it took was something like a minute and a half. Adjustments to this kind
of image really let the power of LAB shine. I tried correcting this image using a variety of
techniques in RGB but simply could not get nearly as good a result. Try as I might, the
evergreen trees never got close to good. Much of the cast was removed easily by dragging down
Saturation in cyan in the Hue/Saturation dialog box, but...it left an extremely annoying and
mysterious cyan fringe around the edge of the road.
Anyway, the bottom line is that you need to know which images are best for LAB. It is not a
"given" that all images should be worked on in LAB. But knowing which ones LAB is well suited
for will make all the difference. For color casts, nothing beats LAB.
1, 2, 3, 4, 5 ( p