Page 2 of 2
Posted: Fri May 22, 2009 5:32 pm
I'll play some more..but here is some food for thought...
I took this into the CM masking world and took a hack at it. I used the K channel mask and limited my adjustment to the B channel as shown. This has to be close to what is happening under the hood of JL. I adjusted the mask vertically first by Pivoting the K channel to a horizontal line then adding the curve. It is over the top a bit but the results are pretty similar. I Can control the B channel colors using the mask created in K. If I set up 3 layers L, A And B as smart filters and apply the K channel mask in this mode It might give similar results...
Posted: Fri May 22, 2009 6:47 pm
Jacob will say whether this is similar - I believe not because the Linear Light blend is a violent overlay mode, when gray (ie the flat curve) does nothing and one then dials in colour as one likes by adding or subtracting over the 'mask'. The other interesting points are
a) the point on all curves is the same for any pixel on the image, regardless of the individual channel values
b) it is possible to set multiple (neutral) points in Lab
If you take the red curves Jag and try changing that colour (using the A channel) - and then play with the zoom, you will probably find that the result is completely different from CM. You can go from pink-ish to blue, just by changing the Zoom opacity! (There is a small problem with the fence colour that one doesn't get using Lab normally)
Posted: Mon Jun 08, 2009 10:47 am
Greg: first, what were you using as an input channel? Second, those a* & b* curves in your original post look pretty extreme to me. What part of the image are they hitting?
The real key for me to color correction using the ladder or any other method is approaching the image with the right mindset. I think these steps as described below generally happen anyway, but it's good to be explicit about it:
1. Clearly state the goal of your adjustment(s). For instance, (1) "I want this sky to be darker and bluer" or (2) "I want to increase the contrast between leaves and branches" or (3) "I want to separate the pink parts of this flower from the red parts, making pink more neutral and red more colorful", or whatever.
2. Figure out how to use the tools you have to meet your goals... Which means figuring out how to map the model of the world assumed by the toolmakers onto the model in your head. I find the ladder much closer to my goals than tools baked into Photoshop. For instance, in the examples above: (1) Using the b* channel as input, bump lightness down at the left end of the curve, and then do the same for b*. Possibly push a* up or down for that region too, to adjust hue towards purple or cyan. (2) Probably use a* as input, or perhaps L*, then click some points onto the curve for branches and leaves, and push them apart in L*, and maybe a bit in a* too. (3) Use a* as input, and make the curve slope upward between pink and red regions, maybe by dropping the pink parts below 0, then push L* up for pinks and down for reds. Etc. Etc.
I'm actually writing some these days, with screenshots, but other tasks keep popping up.
Posted: Mon Jun 08, 2009 10:49 am
Using a mask has a completely different effect. It can be useful, to be sure, but it's not the same concept.
* * *
As for noisy channels: Try using a Gaussian blur or surface blur on the ladder inputs, if they're otherwise creating excessive noise or banding.
* * *
Finally: Yeah, (a) it doesn't seem to be possible to use Photoshop's info palette to get info about the global values until the curve has been put away, and (b) the whole color-shift-in-sub-100%-zooms issue is really obnoxious, a straight-up bug as far as I'm concerned. Possible ways to deal with this include: working on a down-rezzed version of the file and then copying the curves over when finished; working at 100%, or at least keeping a few views of the image open at 100%; Making sure to merge visible layers into a new layer from time to time, to make sure you're doing what you want.
Photoshop CS4 may help with one or both of these, but I haven't tried it.
Posted: Mon Jun 08, 2009 8:26 pm
Thanks for the response. As usual I will have to absorb the response and try out the suggestions...