What the difference between these?

jfriend

Since I've been using LAB mode, I know that it's good to use an L channel curve to control your tonal brightness without impacting color.

Now, when I have a file in RGB mode that I want to do a quick tonal change to with a curve, I make a RGB curve adjustment layer and set it to Luminosity blend mode. I was surprised to see how much difference there was between an RGB curve adjustment layer in normal mode vs. one in Luminosity mode and how cleanly it works in Luminosity blend mode from RGB (no observable color shifts)

Out of this set of observations comes two questions:

1) Is an RGB curve with a luminosity blend mode the same as applying that curve to the L channel in LAB mode? Would the same curve applied in both circumstances give you the same results.

2) Why does an RGB curve affect color?

edgework

jfriend wrote:

Since I've been using LAB mode, I know that it's good to use an L channel curve to control your tonal brightness without impacting color.

Now, when I have a file in RGB mode that I want to do a quick tonal change to with a curve, I make a RGB curve adjustment layer and set it to Luminosity blend mode. I was surprised to see how much difference there was between an RGB curve adjustment layer in normal mode vs. one in Luminosity mode and how cleanly it works in Luminosity blend mode from RGB (no observable color shifts)

Out of this set of observations comes two questions:

1) Is an RGB curve with a luminosity blend mode the same as applying that curve to the L channel in LAB mode? Would the same curve applied in both circumstances give you the same results.

Close. I tried identical curves in two or three files and the lab version has better detail. Also, despite the fact that the RBG curve was in Luminosity mode, I noted a red drift that was not apparent in LAB

jfriend wrote:

2) Why does an RGB curve affect color?

Do you mean in Normal mode? There should not be a cast added, since all three channels are equal in weight and equal values give dead neutral, but a master curve is going to affect hue and saturation as well as value, whereas, in theory, Luminosity mode leaves the first two alone and only affects value. But, for my money, LAB achieves this more cleanly and with better detail.

jfriend

Thanks

edgework wrote:

Close. I tried identical curves in two or three files and the lab version has better detail. Also, despite the fact that the RGB curve was in Luminosity mode, I noted a red drift that was not apparent in LAB

Thanks.

edgework wrote:

Do you mean in Normal mode? There should not be a cast added, since all three channels are equal in weight and equal values give dead neutral, but a master curve is going to affect hue and saturation as well as value, whereas, in theory, Luminosity mode leaves the first two alone and only affects value. But, for my money, LAB achieves this more cleanly and with better detail.

Yes, I did mean normal mode. The specific example I saw was when I tried to darken a foreground object that was too bright (due to being closer to the flash), it turned very red on me so that fits with your observation. When I merely switched the RGB normal mode curve layer to luminosity mode, the color shift disappeared.

The reason I'm asking these is sometimes it's just really inconvenient to go to LAB, particuarly when you have a bunch of adjustment layers that you want to preserve. And, if it's a last minute quick fixup thing (as this foreground fix was something I only noticed after finishing with the rest of the image), sometimes its a lot more convenient to just stay in RGB.

rutt

jfriend wrote:

Thanks.
The reason I'm asking these is sometimes it's just really inconvenient to go to LAB, particuarly when you have a bunch of adjustment layers that you want to preserve. And, if it's a last minute quick fixup thing (as this foreground fix was something I only noticed after finishing with the rest of the image), sometimes its a lot more convenient to just stay in RGB.

You can always duplicate image, flatten, and take the duplicte to LAB to work on, and then bring it back to use as a layer or whatever.

jfriend

Good idea

rutt wrote:

You can always duplicate image, flatten, and take the duplicte to LAB to work on, and then bring it back to use as a layer or whatever.

Yeah, I do that sometimes when there are a lot of layers already in the RGB image. I've been thinking about a couple of actions to help save time. I didn't think of it in this case and the change was so minor and the luminosity blend mode seemed to solve the problem.

edgework

rutt wrote:

You can always duplicate image, flatten, and take the duplicte to LAB to work on, and then bring it back to use as a layer or whatever.

Even easier: merge to a new layer. Duplicate that layer to a new file. Take that into lab and do what you have to do, then bring it back and replace the original duped layer; call it "from LAB" or something. I'll usually save my lab file as well, if I've made some major moves. That allows you to keep all your masks available for either space.

jfriend

That's what I do

edgework wrote:

Even easier: merge to a new layer. Duplicate that layer to a new file. Take that into lab and do what you have to do, then bring it back and replace the original duped layer; call it "from LAB" or something. I'll usually save my lab file as well, if I've made some major moves. That allows you to keep all your masks available for either space.

That is almost EXACTLY what I do when I'm working on an image that I plan to spend a lot of time on. Instead of merge to a new layer, I just do Select All, Copy Merged, Paste into new file, then copy/paste back when done. When I haven't saved my LAB workings, I've sometimes regretted it it later on. The desire to make a small change in the LAB adjustments makes you start over if you don't also save the LAB adjustments file. I wish there was a way to save your LAB workings in the original file, rather than have to keep a separate file.

I've been daydreaming about two JavaScripts to do this for me automatically. The first script would create the new image, name it as a derivative of the first, paste in a merged copy of the first, convert to LAB and set up my default layers. The second script would find the original image based on the derived name of the LAB image and copy the LAB result back to the original as a new layer with an appropriate name. I know you can do most of this with actions, but I don't think you can find the original source image automatically with an action in order to paste the result back.