Most books on Photoshop will tell you that Curves is the way to edit the tonality (and colour) of your images. What they don’t mention is that an edit with the composite RGB curve (the “RGB” channel) performs as you’d expect only for neutral colours (R=G=B). The further you move away from this neutral axis, the greater the distortions are. These distortions manifest themselves as unwanted changes to saturation and even hue.
Let’s look at how the composite RGB curve works. A single RGB colour is constituted by its red, green and blue component values with an equivalent calculated luminosity point on the composite RGB curve (indicated when you click on the image with the Curves dialog open). Moving this point on the composite RGB curve makes identical moves to each component channel around a luminosity value which is unlikely to relate to any of the original component values. The result is a variable change to the component values depending on where they fall on each channel curve and the shape of the curve itself. As the relationship between the component values changes so will the colours change, and not just in luminosity.
Adobe either thinks you won’t notice, or you’ll find the attendant changes to saturation and hue “pleasing, more film-like”. Remember that the reason you’re using Curves (rather than other tools) is so you’ll have more precise and direct control over edits to your image. Others will tell you that setting the Curves adjustment layer to Luminosity blend mode will fix all this, forgetting that it will extract the luminosity of the result (namely after the distortions) rather than apply the Curves directly to the original’s luminosity. While a simple Luminosity blend can prevent hue shifts, crucially it doesn’t preserve the luminosity relationships of the original colours.
What’s really needed is a way to separate out the tonality and colour components of the image and allow edits to each in isolation. One way to do this is with Photoshop’s Lab mode but there are serious downsides to this which are beyond the scope of this article. Another is to use Lobster, the brainchild of Ian Lobb with programming by Michael Cutter, both from Melbourne. (Anybody who has done one of Les Walkling‘s excellent Photoshop courses will be familiar with Lobster.) This is a commercial add-on to Photoshop which leverages Photoshop’s support of luminosity as an intrinsic attribute. You can read about the whys and wherefores of Lobster on Ian’s site and also download a demo version. What I want to do here though is present an approach to using Lobster that I’ve found both practical and very powerful.
When you run Lobster on your image, you’ll end up with a layer structure like this:
Your original image is retained (for mask creation etc) and below are the LUMINOSITY and CHROMATICITY components. CHROMATICITY is everything in your original image that isn’t LUMINOSITY so when you glue the two together you’ll end up with something identical to your original. Lobster also separates out the individual CHROMATICITY channels for individual editing.
For my own use however, I run a Post Lobsterfy action immediately after applying Lobster for the separation which results in the following layer structure:
The original has been discarded (to save storage), the CHROMATICITY components combined (but can still be accessed individually as channels or with Blending Options) and the adjustment layers I typically use for editing added. Here’s how I apply the above:
- Tonal adjustments are applied to the composite RGB curve of the Tonality layer (no side effects as R=G=B). This is pretty straightforward and probably what you thought you were doing with the original’s composite RGB channel. I find this ideal for setting the tonal end points as well which are now visible with the histogram included in the Curves dialog for CS3 (and later). Larger tonal moves applied to the LUMINOSITY component will generally require edits to saturation in concert, but in a more controlled fashion than you get with standard composite curve moves (see below).
- Dodging & burning (if desired) can be done by making visible the Dodge/Burn layer and applying Brush, Gradient tools etc. If more strength is required the opacity can be reset to 100%. Edge burning (to stop your eye from wandering off the page) is just one use for this layer.
- Sharpening (USM or Smart) is applied directly to the LUMINOSITY layer. This avoids smearing of the individual channels which works against the sharpening effect and is pretty unsightly. You will probably find that less sharpening (a lower Amount) than you normally specify is required. The layer can be first converted to a Smart Object if you want to retain sharpening settings. The LUMINOSITY layer is also ideal as a target for Shadow/Highlight adjustments.
- Global saturation moves are applied with an S-curve to the RGB composite channel of the Colour adjustment layer. This curve shape protects the saturation extremes in a similar fashion to CS4′s Vibrance but with more control. The histogram handily shows how much headroom you have in the current working space. (Note that flipping between Normal and Saturation blends for this layer seems to show negligible changes to hue.)
- Colour adjustments can be made by targeting the individual channels of the Colour adjustment layer. The central locked point in each channel represents neutral with the channel colour and its opposite on the colour wheel above/below this midpoint respectively (namely Red/Cyan, Green/Magenta and Blue/Yellow). So if you want to increase saturation/variation in the reds, you’d move only the top half of the curve for the Red channel to the left and for cyans the bottom half to the right. Overall hue changes are done by moving the midpoint in the desired direction. All this sounds a bit complicated but if you’re familiar with a/b channel edits in Lab mode, it’s pretty much the same except you now have three channels to play with.
The use of Lobster with the above layer structure and approach to editing supports most of the things I need to do to an image to knock it into shape. All edits are preserved in the layer stack so I can continually refine the results without degrading the image. Give it a go … you may be surprised how powerful this can be!