Here’s a doosie. This image is dismal, normally a throw away shot that should never be shown to anyone. Ever. Under any circumstance.
I love New York, I really do, and, well, when I took this pict… I mean, when the anonymous photographer took this picture, he or she wasn’t really sure if it the little point and shoot camera was considered an “approved electronic device” and, well, at this point of approach, none are. So it’s barely in focus, the settings on the camera are what they were for the last picture taken, there was only time for one or two shots, so the composition is.. well.. there’s a huge wing in frame. But it’s NYC and since the pilot probably wouldn’t have taken kindly to a request to loop back around for a better shot, this is what we’ve got to work with. It’s a lousy picture, but Lab colorspace enhancements really rock with pictures like this: unsaturated, hazy, snoozy. Let’s see what we can do with it.
You can download the original from flickr by clicking here to play along with the image if you want. The best way to get comfortable with Lab maneuvers is to play with them. Once opened in Photoshop (the images below are from CS3 but should be similar in CS2), convert the image to Lab mode (Image -> Mode -> Lab Color). I’m only showing a cropped section of the whole image to show more detail. Here’s the starting version:
The first change I made was a Levels adjustment in the Lightness channel (Layers -> New Adjustment Layer -> Levels…). There are three sliders right under the histogram that can be repositioned to set the black point, gray point, and white point.
The far left slider represents the black point (and is conveniently filled in black). I slid this over to 48. This setting cut into the mass of the histogram a little, which is generally bad as it causes clipping, but shadow clipping is a lesser offense than highlight clipping and this image was so bad to start with, it didn’t really hurt much. The far right slider (filled in white) sets the white point and I slid it over to 238 (it started at 255) careful not to cause substantial clipping in this case. The gray point slider behaves differently and adjusts the overall lightness. Sliding it to the left will lighten the image, to the right will darken it. I slid it to the right to 0.90 which looked best to my eye.
It’s mildly interesting to note that changing the levels on the Lightness channel are not actually setting the white and black points of the overall image as they would be in an RGB channel levels adjustment. They are setting the points of maximum and minimum lightness for the overall image. In reality, they are actually setting the white and black points for the Lightness channel which looks like a black and white version of the image. This is just a minor point that has probably overly complicated the discussion, so I’ll stop while I’m behind. Here’s what the image looks like after the Levels change:
At this point, the image is already looking a littler better, but the real magic of Lab adjustments come from Curves. In this case, I didn’t flatten the image at this point so that I could go back and tweak the Levels settings later if I wanted to. Create a new Curves adjustment layer (Layer -> New Adjustment Layer -> Curves…). Here are the Curves settings I ended up with (and the unsharp mask settings which I’ll get to later) — I’ll discuss each in turn.
The top right “a” channel curve is a simple steepening of the curve, like turning up the gain. The curve passes through the origin (center point) of the graph so there is no shift in the color cast in this channel. Since the “a” channel represents magenta to green colors, these will be intensified a bit by this curve. The points I chose were +/- 77 (as you grab the little squares in the corners and move them around the curves area, the Input value will change).
A little more interesting magic happens in top left “b” channel curve. In this case, we don’t pass through the origin, instead passing a good bit to the right of center. I could have done some white balancing in a number of ways with this image, but I chose to remove the dingy yellow cast in the original image through the Lab “b” curve. A good way to get familiar with Lab and an excellent way to remove color casts using it is to observe the Info helper (Window -> Info or F8) as you move the mouse over the image. For the pixel currently under the mouse, the Info helper will show the L, a, and b values. A color neutral value anywhere from white through gray to black should have a and b values of zero. If you identify some elements in the image that should be white or true gray (in this case I chose parts of the wing that I didn’t feel were reflecting blue sky), mouse over them and observe the a and b values. In this case, the b values were consistently around 8 when I thought they should have been 0 and the a values were close to 0. I ended up with values of -77 and +92 for the endpoints of the b curve.
The Lightness channel curve is generally used to add some contrast to the image and I like to change it after making the color corrections. In this case, I used a fairly subtle contrast enhancement S-curve. The premise here is to increase the slope of the curve where you want additional contrast. There will be better images that we’ll examine in the future to show how to judge this adjustment, so we’ll patiently wait for now. Clicking OK saves the Curves settings and we see:
Without any exceptions (only when there are exceptions), all images deserve sharpening. This is to say that most images look better after they’ve been sharpened. A better (and less sarcastic) absolute is that no image should ever, ever, under any circumstance (well, maybe a few interesting ones that we’ll also discuss in the future) should sharpening be applied to color channels, only Lightness or Luminosity. If you are in Lab already, it makes sense to apply the sharpening to the L channel.
Press command-1 (on a Mac, or Ctrl-1 on a PC?) and only the L channel will be displayed, looking much like a black and white version of the image. Anything you do with Photoshop at this point will only be applied to the L channel. In this case we want to use Unsharp Mask (Filter -> Sharpen -> Unsharp Mask). I used some mild sharpening settings as shown in the lower right of the figure above. After OKing these settings, pressing Command-` (Ctrl-` on PC?) brings back the composite image. At this point all that is left is to switch back to RGB, or CMYK if that’s what we need to do (Image -> Mode -> RGB) and we’ve got our final image:
It’s still not a great photo, but I would definitely not be nearly as devastated showing someone the final version. Lab color enhancements are not appropriate for every image, but this is an example of one in which Lab really shines. Images that already have a lot of saturation or vibrant colors are not going to be helped by the steeper a and b channel move here and may produce colors that can’t be printed or viewed on a monitor. But if you have a desaturated, color starved image, this simple enhancement process can make a huge difference. If you play along with the image in Photoshop yourself, the differences in the image at each step really pop, but seeing them one at a time and having to scroll a window up and down to see the different stages don’t do it justice. So in parting, here’s a before and after composite of this enhancement: