View Full Version : L.A.B. Color Correction Tutorial by John Chardine

Lance Peters
08-21-2010, 11:59 PM

Lance's original post:

A cleaner BG for a big change.

Sigmonster @800mm with 36 extension tubes
1600 ISO
Matrix Metering
Braced on car window

Cropped and Sharpened.

What could I have done here to improve this shot??

Myer Bornstein
08-22-2010, 03:38 AM
Maybe a slight more head angle:D

Joel Eade
08-22-2010, 07:10 AM
Hi Lance,

Beautiful bird, I like the background and the eye, on my monitor it could stand another touch of sharpening and I would like to see just a little more space on the left.


Jeff Cashdollar
08-22-2010, 07:24 AM
Like the composition and great eye. The f/7 is a sweet spot for lens as well.

Julie Kenward
08-22-2010, 07:47 AM
HA is the first thing I noticed, followed by the white patch right above his beak - is it hot? I can't quite tell...

I also think a little more space left would be nice and another round of sharpening on those body feathers wouldn't hurt.

If you wanted to get REALLY picky, you could blend the white horizontal band at the bottom of the image with the rest of the BG. ;)

Bill Dix
08-22-2010, 08:29 AM
Looks awfully good to me, including the HA. Slightly more space to the left would position the eye closer to the UR-ROT spot, and maybe a touch of sharpening as noted. On my screen there seems to be a bit of bluish cast in the shaded areas; common enough, but I wonder what a tiny bit of orange filter at low opacity would do to the image? Since no one else has mentioned it, I'm probably all wet. Ditto Jules on the white band.

John Chardine
08-22-2010, 12:30 PM
I love this species Lance!

Got it in one Bill- big bluish cast on my calibrated monitor too. I use the LAB colour space to correct casts, sharpen, reduce noise, saturate or desaturate etc.

The OP was running +4 on the A channel of LAB and a whopping -20 on the B. The B channel runs from Yellow (above zero, positive numbers) to blue (below zero, negative numbers), so there is the bluish cast staring you in the face. I converted image to LAB and pulled up or down the Curves on A and B to approximate A = 0, B = 0 on neutral tones (whites, greys). This is the result.

PS I can go into more detail on the method if needed.

Michael Gerald-Yamasaki
08-22-2010, 01:08 PM

Greetings. I wonder a bit about the color, the whites (?) seem quite a bit on the blue side. Not familiar with this bird (so I have no idea what the colors "should" look like)... I messed with the color a bit, trying to enhance the violet-green contrast (please excuse the quick masking job)... just an idea - only color adjustments:


Eyedropper wb in LR on about 1/2 way to the right of the eye. Curves in both RGB and a wide gamut CMYK mostly Blue, Yellow, Cyan, and a touch Magenta channel adjustments. Multiply blend for bg.

Just saw John's take... interesting blue differences.



Lance Peters
08-22-2010, 03:58 PM
John - nailed the repost and Congrats Bill for getting the colour cast.
Michael - you seem to have given him a purplish cast around the back and under the bill area.

Would have liked a little more head turn towards me.

Well done!

John Chardine
08-23-2010, 07:24 PM
Lance asked me to go into the method I use to colour correct in LAB colour.

First, there are various colour spaces or modes within which you can work in Photoshop. The common one is RGB which has a Red, Green, and Blue channel and various combinations of each make up the gamut of colours available in the colour space. Digital cameras have sensors made up of RGB sensor sites and computer monitors have RGB pixels so RGB is a convenient place to work. In RGB, the detail in an image and the colour information are encoded in each of the colour channels- each one looks like a different black and white version of the image. Check this out by opening an image and going to the Layers palette and clicking the Channels tab, then click in turn the red, green and blue channels.

The LAB (pronounced El Eh Be) colour space has three channels also- L for Luminance, A for a colour range from Magenta to Green, and B for a colour range from Yellow to Blue. This is the toughest part of LAB- getting your head around these two colour ranges. What does blue and yellow have to do with each other anyway? Amazingly, even though there are only two colour channels in LAB, the LAB colour gamut is a lot wider than in RGB. The big difference between LAB and RGB is that in LAB the detail of an image is contained in one channel only- L, and the colour information is contained in A and B. So detail is separated from colour and can therefore be adjusted separately. The attached image shows Lance's beautiful photograph as it looks in the three LAB channels. Note that the A and B channels don't seem to have much contrast or detail but the L channel is full of detail. Separating detail from colour is why LAB is so powerful for correcting colour casts, sharpening (the L channel), removing noise (in the A and B channels), enhancing and separating colours in an image, making selections, and so on.

To work in LAB mode, open an image and select Image->Mode->Lab color. Then take a look at the channels and have some fun with it. Try adjusting the Curves of each channel (more on this later).

I'm going to leave it there for now and recommend that you do some reading about LAB on the web. Wiki has a good article but it's too technical for me. I also highly recommend Dan Margulis' book "Photoshop LAB Color: The Canyon Conundrum and Other Adventures in the Most Powerful Colorspace", published by Peachpit. Some say it's the most ground-breaking book ever written about Photoshop, and there's been lots!

Next I'll talk about how a color cast is diagnosed and corrected in LAB.

Doug Brown
08-23-2010, 11:09 PM
I'm really enjoying this discussion!

Lance Peters
08-23-2010, 11:56 PM
Great Info John!
Looking forward to more - very educational.

Mark Young
08-24-2010, 04:03 PM
John's repost looks great, though your adult bird now looks like a juvenile bird.

John Chardine
08-24-2010, 08:12 PM
So I want to quickly get into diagnosing a colour cast using the LAB colour space.

Every colour in the LAB space can be defined by three numbers, one for L, one for A and one for B. This is the same as RGB where for example a tone of red might be R=238, G= 13, B=13. In LAB, the L-value gives the brightness or Luminance of the pixel from 0 (black) to 100 (white). The A-value gives a colour from Magenta to Green, and the B-value gives a colour from Yellow to Blue. Neutral tones (i.e., those with no colour) are coded A=0, B=0. That goes for black, white and all the neutral gray tones in between. Numbers on the plus side of A code for increasing Magenta (a warm colour), and on the minus side code for increasing Green (a cool colour). Numbers on the plus side of B code for increasing Yellow (a warm colour), and on the minus side code for increasing Blue (a cool colour). So that same red I mentioned above in LAB works out to be L=59, A=86, B=73. This means it's made up of lots of Magenta and lots of Yellow. A Blue I looked at in LAB worked out to be 56,-19,-52 for L, A, and B, so in this case some Green and lots of blue.

So again neutral tones in LAB get values of A=0, B=0. This is the basis of diagnosing a colour cast in LAB. To look at LAB values in parts of an image, open up the Info panel and pull down the small menu, top right and choose Panel options, and select LAB color for the Second color readout. Then mouse around an image and see how the LAB values change. In the attachment I show Lance's original image. I've placed the cursor over a portion of the image that I judge should be neutral- in this case some white feathers in the shade. The Info panel shows that the pixels under the cursor have values of A=4 and B=-20, which means a little on the Magenta side of neutral in A and heavily Blue in B. Again, if those feathers were truly neutral they should be reading close to A=0, B=0.

So that's how you diagnose a colour cast in the LAB colour space. The process is somewhat subjective because you have to decide what parts of the image "should" be neutral and then proceed from there. In warm, evening or morning light the whites and greys are going to look warm and the A and B values in LAB will both be positive. You would not necessarily want to correct for this because you would remove the very effect you are trying to achieve by going out early or staying up late.

In the final installment, I'll run through the method of correcting a colour cast in LAB once diagnosed.

John Chardine
08-25-2010, 08:25 PM
We have reached the stage where we can actually correct a colour cast in the LAB colour space.

Step 1: Convert your image temporarily to the LAB colour space or mode. Choose Image->Mode->Lab color. Photoshop does a pretty good job of converting and you will not see any difference in the image at this stage.

Step 2: Open up the Info panel so you can see the LAB colour readout as you move the eyedropper over the image.

Step 3: Now here's the fun part- chose Image->Adjustments-> Curves or cmd/crl-m to bring up the Curves window. Drag the window around the screen so that you can see it, the image and the Info panel all at once.

Step 4. You will notice that the Curves window has a pull-down menu called Channel and because you have converted your image to LAB, you have Lightness (L, Luminance), a and b channels (Adobe insists on calling the channels this instead of just L, A, and B). Our diagnosis of Lance's image suggested that neutral tones were running about -20 in b and about +4 in a, so we know b is farther away from 0 than a is. Let's correct b first.

Step 5. Choose the b channel from the pull-down menu in Curves. Click and hold the mouse while moving over the image and you will see a small circle moving over the histogram which shows you where the pixels under the eyedropper are found. Mouse over the neutral-tone we previously diagnosed on the neck of the bird and check out where that is on the histogram. To correct the colour cast caused by the B channel, click and hold the diagonal line running through the histogram at the point where the neutral tone pixels reside and pull the line up or down. The direction depends on the direction of the colour shift in B (+ or -) and if you have set the histogram to show white on the right or the left. Don't worry, just pull the line up or down, let it go and check out the Info panel as you mouse over the neutral tone of interest. Have a look at the attached. Here, I've pulled the line down and it has changed the neutral tone from -20 to -1, so down was the right direction. If I had pulled the line up I would have made the problem worse (say -20 to -30). The Info panel is neat because it shows you the before and after values for the part of the image under the eyedropper. Pull down a little bit more to get B as close to 0 as possible for the neutral tone. Repeat for the A channel and there you have it, your image is corrected so that neutral tones really are neutral, that is no colour, A=0, B=0. It's quite alright to have neutral tones running -2, -1, 1, 2 etc so don't get too hung up on 0,0.

You can do some nifty things with the Lightness/Luminance channel in Curves but we'll leave that to later. I would encourage you to give this a try and if you have problems, post queries here.

Michael Gerald-Yamasaki
08-25-2010, 11:28 PM

Greetings. Thanks much for the explanation of LAB... I find working in other color spaces than RGB a rewarding experience, but haven't explored LAB much. One thing I noticed while reading over your posts and working this image some more in LAB is that LAB seems to be a good mode to work color contrast. I noticed in the curve you posted above that the curve being more horizontal in the part of the histogram where most of the pixels are would mean a smaller range of color (the resulting histogram would have a thinner peak). Using a different, steeper curve (while still balancing the color cast) would both correct the cast and add color contrast (here along the yellow-blue axis of the B channel).


(Note the change in the histogram from the Pigment/Ink % to Light. Sorry, the histogram will be inverted in comparison to yours, but the slope differences will be similar). Note also that the sampler point values displayed (on the far right) show the change of the b channel for that point from -13 to 0 (as you described removing the cast). Not shown is the slight adjustment to the a channel.

If you try this for yourself, take a look at the differences in the b channel display after applying a curve as in this post.

Anyway the result is more a matter of taste, but I think a pretty neat way to adjust color contrast to one's liking...



John Chardine
08-26-2010, 10:17 AM
Thanks Michael. There are several ways to skin a cat in Photoshop (apologies to cat lovers!).

I like the idea of pinning some sample points on the image. That way you can see the results of adjusting the curves in the Info panel without mousing over the image. To place a sample point on the image shift-click the eyedropper tool. You can place up to 4 points at least in CS2 which I have running in front of me right now. Shift-opt/alt click to remove them.

The specific method you used to adjust the image accomplished two things- removing the colour cast and upping the saturation and colour separation in the image (you call this color contrast, which I guess is the same thing). I prefer to make one change at a time and you can accomplish this with the method you describe by leaving the diagonal line at 0,0 bottom left and only bringing in the top right point. With the way you have the curve display settings showing Light instead of Pigment/ink %, that move will shift the B channel towards the Yellow and away from the Blue. With the way I have it set (showing Pigment/ink %, as per Dan Margulis' recommendation) I would do the reverse- slide in the bottom left corner and leave the top right corner the way it is.

Increasing the the slope of the Curves diagonal in the A and B channels is a major theme in the Margulis book, and is the way he recommends to enhance and separate colours in flat images out of the camera such as those commonly produced in the canyons of the sw. US- the "Canyon Conundrum".