As I understand it, moving the midtone input slider (aka gamma) in photoshop levels adjustment, causes highlights or shadows to compress. If I have an image with lots of red in the highlights, for example, does this mean that using the midtone slider may change more than just brightness? Would it change the relations between colors and hence induce a color/tone shift? When I use exposure in camera raw or the new brightness adjustment in photoshop (cs3 and up) I can get similar results but without a perceived color change in the highlights. What is the correct way to brighten an image and keep the overall color of it? I had an image of a model and I thought I saw the skin tone and maybe even the white balance change when using the levels and midtone technique.

I would appreciate your help,
Yakov

If all you're sliding is the midtone input then you are not changing either the highlights or the shadows, only the stuff in between, a.k.a. the gamma. It should not change your color balance at all, just he levels.

As I understand it using the midtone slider in a level adjustment layer, will as you say compress either the highlights or the shadows, depending on direction. It is really the same thing as using a curve layer and pulling the curve up or down with a point in the middle of it. I you move the midtones, the other tones (highlights and shadows) need to adjust, don't they? :-)

As I understand it it should keep the relations between the colors, I have not noticed color shifts myself. But I'm not sure.

http://en.wikipedia.org/wiki/Gamma_correction

Continuing with this. If you make large curve adjustments, it can look as colors shift sometimes. If this is due to that the actual balance between the colors shift or that a color tone is perceived as different at different luminosities, I do not know...

And with this in mind, I guess that a large shift of the midtone slider in levels, can give shifts in the perceived colors...

Back to the original question...
I guess you will get as many answers as there are photographers... :-)

One technique I use in final processing if I want to brighten the image a little is:

- merge all existing layer into a new layer
- set the blend mode to "Screen"
- Adjust the opacity to how much you want to brighten the image, often i use values around 10-40%

My understanding is that 'colour' (by definition of 'colour' in strict sense of the word) should be preserved for any adjustment which preserves H and S channel (i.e. only L channel changes) for each pixel in the image (in HSL colour view), note that if you change L all R, G and B will change (in RGB colour view), that is because RGB contains not only 'colour' in strict sense of the word but also luminance (L cannel or light intensity), say R=255,G=255,B=255 (in 8-bit RGB colour space) is (what we call) white, while say R=200,G=200,B=200 is (what we call) grey, these two colours are in fact the same 'colour' ('shades of white' so to speak) and they differ only in L (have the same H and S). I would assume that levels, curves, contrast, brightness, etc. adjustments in principle affect only L channel of each pixel in the image (but I'm not sure).

I guess that non-uniform change of L cannel across the pixels may give impression that the actual 'colour' has changed, and in loose sense of the word it really did change (as in these terms the 'colour' does include luminance as well) but in strict sense of the word the 'colour' did not change (H and S remained unchanged), say if you darken 'reds' (red-like pixels) in an image (decreased L for 'reds') you may get a feeling that the hue and saturation of the 'reds' did change while in fact it did not.

http://en.wikipedia.org/wiki/HSL_and_HSV
http://en.wikipedia.org/wiki/RGB_color_model

It is possible that I've made a mistake in my previous post (sorry can't edit the original post)...

when I said that...

say R=255,G=255,B=255 (in 8-bit RGB colour space) is (what we call) white, while say R=200,G=200,B=200 is (what we call) grey, these two colours are in fact the same 'colour' ('shades of white' so to speak) and they differ only in L (have the same H and S).

...I was most likely wrong in assuming that the two 'colours' have the same H and S, they (definitely) have the same H but (very likely) different S, therefore (according to my definition of 'colour' in strict sense of the word) they are not the same 'colours'... sorry about that (all the rest should hold)...

Thank you all for commenting. I am not sure I am less confused now. Adobe states that moving the midtone slider compresses one side and spreads the other. So if one side has more red, you should get a more pronounced effect of compression on the reds in the image.

A followup question, I get different results when I lighten an image using the highlights marker (moving it towards the middle and not clipping channels) and when I use the brightness adjustment. The brightness adjustment does something similar to a curve. I can check the color of the skin with CMYK values to find out that in some areas the skin tone changes between lightening methods. Using two layers with screen blending shows a different behavior. So which method will brighten the image and keep the same skin color? Some changes can be seen as a the skin turns towards the red. Working in Lab mode gives yet another result, unlike any other. All will cause the image to lighten but all with different effects. It is obvious to me that not only luminous is being changed in some methods. All methods differ in color (perceived and CMYK and RGB values) and in overall image contrast, so I guess each method does something a bit different to the image.

Sorry if I sound confused (I am) but I want to understand what it is exactly that I change when I use those different methods for lightening an image.

Thanks again for your participation,
Yakov

In my opinion, in simple terms, most of image adjustments, such as curves, brightness and contrast adjustment, 'midtone slider' adjustments, etc. should change the colour (even) in the loose sense of the word (i.e. hue or saturation would change), if you modify either R, G or B channel (or in any way all of them) you have changed the colour in the strict sense of the word.

Maybe in order to affect the colours as less as possible (i.e. to change luminance only and leave hue and satiation unchanged) one could decompose the image in H, S and L channels and then apply the changes (say mid point slider, or curves, etc.) only to L cannel (not sure if Photosho allows you to do that).

Looks like in Lab colour mode in Photoshop you can apply various adjustments only in luminance channel (below is given a sharpening example in "1. Sharpening" section)...

http://www.luminous-landscape.com/tutorials/instant_photoshop.shtml

That's what I usually do when processing, and it does make a difference