Untitled Document

On page 292 of Dan's Lab book it describes using the Channel Mixer to help the vegitation in RGB before moving into Lab for further corrections. My problem is that I can't seem to wrap my head around the math of the Channel Mixer.

I understand why (on page 292-293) the Green is moved to +130% (after all we want to open up the Green channel), but the next move of Blue at -30% has me stumped. I realize that the move brings all previously balanced areas back into balance, but why not move Red -30% instead of the Blue? After doing it I see the results aren't as good (not really good at all), but I can't seem to explain to myself why.

I am sure the problem is that I am very familiar with CMYK and Lab, but moves in RGB aren't as easy for me to grasp. There, I admit it. And when someone helps push me past this Channel Mixer hump it will hopefully wipe away the RGB cobwebs.

Andy Adams

___________________________________________________________________________

The answer can be found in the following link :

http: //www.ledet.com/margulis/2007HTM/ACT06-3_shades_of_green.htm

In particular, the following explanation from Dan :-

"1) Starting with an RGB file with poor greens but no obvious color cast, on a duplicate layer (not an adjustment layer) Image: Adjustments>Channel Mixer. Choose Greens and enter Green +140%, Blue -40%.

Explanation: The procedure of subtracting part of a channel from another is usually not advisable because it costs detail in darker areas (the darkest areas of the channel are subtracted more heavily than the lighter parts). In natural greenery, however, the blue channel is usually almost solid, so subtracting doesn't harm detail in the target channel."

In fact, Dan's also mentioned the reason of choosing Blue in the first para. of page 292 of the Lab book "...a blue channel that, like Figure 14.5C, is dark and lacking in detail in the green areas."

If you still don't understand, may be Dan can explain this to you.

Wai-hong Chung from Hong Kong

___________________________________________________________________________

Andy, I would also like to know why anyone would choose to use a blunt instrument like Channel Mixer when the much more precise Curves tool is available. For example, in enhancing greens Dan recommends ( in another of his books) adding magenta to green to improve contrast by darkening the green shadows. With Curves it's a very simple matter to add magenta anywhere you want it-highlights, midtones, or shadows. Using the Channel Mixer as described on p. 292 left my image with a slight yellow cast. The cast probably could be removed by further adjustments in Channel Mixer, but using Curves didn't take any longer, and it produced a clean image with white clouds, blue water, and green trees. Anyone who avoids using Curves for fear of being harmed should rest easy. I've yet to hear of an instance of Curves jumping out of the shadows and biting or bludgeoning anyone. Fool around with them for awhile and you'll never be tempted to look back. As for understanding the mathematics behind Photoshop's tool, it's a good intellectual exercise but of little practical value. Just move the sliders around and see what happens. With practice you'll get a feel for what's happening and you won't need a calculator at your side. As another example of working without bothering to understand the underlying theory, I've never felt a need to try to memorize the reasoning behind the Shadow/Highlight sliders. Thirty seconds of moving the sliders around usually gives me the results I want when working with my own challenging images. Saving an additional fifteen seconds does not make a good argument for trying to understand and remember all the formal details.

Some of those who read these posts may not be familiar with the fact that RGB curves can be used to adjust not only R, G, and B, but C, M, and Y as well. The latter three are complements of the first three.

Moving a curve up or to the left increases the primary color. Moving it down or to the right increases its complement. (Or at least it works that way when you have your Curves gradient bar set so that the light end is at the top and the right; if you use the opposite orientation, moving the curve up increases the complement and moving it down increases the primary color).

Contrary to the above comments, I should add that I'm as anxious as Andy to understand the workings of Channel Mixer. One can never know too much.

Howard Smith

___________________________________________________________________________

Sometimes when you want to chop down a tree an ax works better than a scalpel. Curves work great when the basic shape is already in each channel, but if there's nothing to work with you need to get that detail from somewhere.

That's where typically Dan would do some sort of Apply Image command, but others would use Channel Mixer in an adjustment layer. Same basic idea, just different angles.

I remember a job a few weeks ago where we had to change the color of some brightly colored beams of light. It's hard to make a yellow beam of light look purple or an orange beam of light look green without Channel Mixer or some variant.

Using Channel Mixer in RGB is much more difficult for me than in CMYK, but I find that using it forces me to think about the channels involved, which is beneficial for any correction and a good exercise.

--

J Walton

___________________________________________________________________________

Howard writes,

Curves are the method of choice if there is a cast in the original image. Commonly, however, even if there is no cast, we or the client decide that the greens simply aren't intense enough. Trying to make them artificially greener with curves often doesn't work because it can impart a green cast to the entire image.

Photoshop offers a gang of other ways to achieve this: LAB, Selective Color, Camera Raw, Hue/Saturation coming immediately to mind. All of these methods fail, though, when the "green" area in the original is so dull that Photoshop has a hard time realizing that it *is* green.

In that case, the Channel Mixer move can be uniquely valuable because no matter how dull the green is, the blue channel is going to be darker than the green. So adding X% of green and subtracting the same X% of blue always makes those areas greener and then we have a good shot at finalizing them, if we wish, with one of the methods described above.

This Channel Mixer method can't impart a cast as such but it can give that impression. At one point I noted the solution but it doesn't seem to be in the link posted by Wai-hong, so, let me repeat it here.

If you use this Channel Mixer move to augment greens, you may, or may not, encounter any or all of the following problems:

1) Blues become objectionably purple.

2) Greens, although greener as desired, are now too light.

3) Browns become objectionably yellow.

The solution is to do the Channel Mixer move on a standard, not an adjustment layer (one can't easily solve problems 1 and 3 with an adjustment layer). Then,

*If you have problem #1, Edit: Fade>Lighten.

*If you have problem #2, change mode of the Channel Mixer layer from

Normal to Color.

*If you have problem #3, convert to LAB and use Blend If to exclude anything

that is positive in the A channel.

I've just had good luck with this in a series of pictures with dark, tree-covered hills in the background, nearly gray, I would like to see them greener and lighter. The sky above them is properly blue but I wish it was a little darker. When I use the Channel Mixer as described above, I get problem #1, but problems #2 and #3 are absent.

Solution: apply the Channel Mixer move on a layer, then duplicate it on a third layer (again, not an adjustment layer). Change mode of the middle layer to Lighten and the top layer to Luminosity. The middle layer makes the hills greener and lighter; the top layer does nothing to the greens, but darkens the non-neutral areas of the sky without changing its color.

Dan Margulis

___________________________________________________________________________

Thanks, Jay! In just a few words you cleared up a big mystery for me. It's embarassing to admit that the dignificance of the name "Channel Mixer" completely eluded me. I was so blinded by the colors that it never occurred to me that Channel Mixer is doing just what it says it will. Anyone who wants to see this in action can do so by opening an image and choosing Window/Arrange new window for.(image title). Make the target channel visible (green in this case) and watch the changes in the Green channel when applying Channel Mixer to the full-color image. Fascinating! And it explains why Russell Brown and others are so fond of using Channel Mixer for creating B/W images from color images. Your comment ".if there's nothing to work with you need to get that detail from somewhere" cleared it all up for me. Now I've got a whole new range of things to look into. Channel Mixer offers great potential for experimentation.

Yours was the kind of answer that I hoped to receive. Thanks for taking the time to help.

Howard Smith

___________________________________________________________________________

Dan, you and J Walton have taught me more about the Channel Mixer than I have been able to find in an extensive collection of Photoshop books. None of the tutorials found on the Net in a quick search come close to providing such useful information. I'm glad Andy brought this up in the first place.

Thank you for taking time to offer such a valuable collection of Channel Mixer information. You really should consider writing another book. Your last four have been so useful that they've been reduced to stacks of single pages, the bindings having long ago given up.

Howard Smith

___________________________________________________________________________

Howard Smith wrote:

A note to all list members (who may use this group via email and not via a web browser), if you go to this groups website, log-in and do a search of the ACTL group (not to be confused with searching other groups which appears first) using the keywords "channel mixer" - you will also find a lot of useful information (if you have the patience to dig through it all)!

http://tech.groups.yahoo.com/group/colortheory/

Regards,

Stephen Marsh.

___________________________________________________________________________

I strongly support Howard's suggestion. Since the issue of PP5E, Dan has updated and added new tricks and I'm sure he has more to come. It would be great if Dan would group these tricks together in a new book.

Wai-hong Chung from Hong Kong