RetouchPRO

Go Back   RetouchPRO > Technique > Photo Retouching
Register Blogs FAQ Site Nav Search Today's Posts Mark Forums Read


Photo Retouching "Improving" photos, post-production, correction, etc.

L*a*b* question: why can't it be simulated in RGB?

Reply
 
Thread Tools
  #21  
Old 09-08-2010, 02:38 PM
andrewrodney's Avatar
andrewrodney andrewrodney is offline
Senior Member
 
Join Date: May 2010
Location: Santa Fe
Posts: 1,090
Re: L*a*b* question: why can't it be simulated in

Quote:
Originally Posted by Chain View Post
And if you want the RGB version to not shift the color the curves (dodge) layer should be set to Luminosity blending.
I too will look at the files (thanks) but the idea mentioned above is what I would have initially thought of too; using the Luminosity blend mode to avoid the color change.
Reply With Quote top
  #22  
Old 09-08-2010, 02:42 PM
mistermonday's Avatar
mistermonday mistermonday is offline
Moderator
Patron
 
Join Date: Dec 2005
Posts: 3,028
Re: L*a*b* question: why can't it be simulated in

Chain, setting the curves' blend modes to luminosity doesn't prevent the shifts nor does the correction work the same.
Regards, Murray
Reply With Quote top
  #23  
Old 09-08-2010, 02:58 PM
Chain's Avatar
Chain Chain is offline
Senior Member
 
Join Date: May 2009
Location: Oslo, Norway
Posts: 551
Re: L*a*b* question: why can't it be simulated in

mistermonday:
I have looked at those files and I found the test a bit faulty (or did i misunderstand the purpose?). I noted some of my issues with it in the edits of the above post.

That said, it was very easy in both LAB and RGB to match the brightness of the two colors using a gentle curves adjustment (brightening/dodging).
* In RGB the color did not visibly shift if blending was set to Luminosity.
* In LAB the color did shift (a bit more purple).
Note: when using Normal blending the RGB did shift the color more.

Attached are my improved files with my results so you can have a closer look.
Attached Files
File Type: zip LAB and RGB test.zip (30.8 KB, 15 views)
Reply With Quote top
  #24  
Old 09-08-2010, 04:19 PM
Chain's Avatar
Chain Chain is offline
Senior Member
 
Join Date: May 2009
Location: Oslo, Norway
Posts: 551
Re: L*a*b* question: why can't it be simulated in

I'm still trying to finding a useful use for LAB worth the flattening/multiple conversions.

For adjusting brightness or adjusting brightness/color separately it does not perform as good as RGB w/blending modes.

Still, I will have to research it a bit more but it seems it might prove an interesting alternate way of playing with color and doing heavy color saturation.
Although with Vibrance, Hue/Saturation (allowing you to selectively adjust different color hues) combined with saturation/hue/color blending modes and all the other tools for adjusting color I think that if it has an advantage it would be in very special cases... For some artistic expression it could be fun to play with either way...
Reply With Quote top
  #25  
Old 09-08-2010, 05:06 PM
mistermonday's Avatar
mistermonday mistermonday is offline
Moderator
Patron
 
Join Date: Dec 2005
Posts: 3,028
Re: L*a*b* question: why can't it be simulated in

Quote:
And if you want the RGB version to not shift the color the curves (dodge) layer should be set to Luminosity blending.
Chain, I may be missing something, otherwise no way. What I would like you to do in the RGB image is to change both eyedropper points in your info palette from RGB to LAB (not the image just the eyedropper samples), all layers on except the color correct layer, make active the visibility layer, and now look at the L values. Both are 58. Now set the Curve blend mode to Luminosity. Make active the Visibility layer and you will see that the Dodged area has changed value to 56. On my monitor there is a very significant change in the lightness of the two gray areas. Now turn the visibility layer OFF and turn on the recolor layer. On my monitor there is a very significant difference in color between the dodged area and the left side of the image. When you are dodging and burning a portrait, a difference in a few points of lightness is extremely noticeable and in most cases an unsuccessful dodge or burn. Shifts in saturation caused by D&B curves must be corrected as they are very noticeable. Small color differences are also not acceptable.
Now play with the LAB file. Feel free to also change the curvve blend mode to Luminosity - it has no affect whatsoever. In LAB, you adjust the luminosity and when you have it right, you can recolor with any color including this case where you go from 11/-58 (A/B) to +38/46 with zero change in luminosity 54/54 and a perfect color match.
While this example is extreme, the principal functioning and implementation is exactly the same for a portrait D&B where the lightness and color differences are equally noticeable.
Regards, Murray
Reply With Quote top
  #26  
Old 09-09-2010, 05:09 AM
Chain's Avatar
Chain Chain is offline
Senior Member
 
Join Date: May 2009
Location: Oslo, Norway
Posts: 551
Re: L*a*b* question: why can't it be simulated in

In this post are you referring to your original images, or my fixed versions? The test seemed a bit flawed to me (especially the original).

Yes, in LAB when adjusting the Lightness channel the a/b channels are not affected and vice versa. And in RGB if i change the red channel it does not affect the blue channel. And...? My point is that the Lightness channel is not a good analoge for the apparent brightness of the image, and the L-value can thus not be trusted to give good enough feedback on the brightness when comparing two different colors. This has to be done visually or with appropriate weighting of the RGB components.
Two colors might have the same L-value but different brightness.

I'm assuming that the purpose of your "dodging" was to increase apparent brightness without affecting color. In that case Luminosity blending mode in RGB triumphs over the L-channel in LAB.

In the test (my version) you can see that adjusting lightness in LAB shifts the blue color slightly towards more purple while in RGB using Luminosity it does not. However it is a bit tricky to compare the results and check them in a fair manner because you have to convert between the color spaces to do this, and also in LAB the Luminosity blending mode does not work.

To digress a bit; that the blue-purple shift is a known weakness in LAB and I believe it is for the same reason that you see the effect when converting out-of-gamut blue tones from RGB to CMYK. The rendering intent tries to move the blue colors into gamut (apparently in the LAB color space) – causing an unintended hue shift. Soft-proofing and manually tweaking the problematic blues is the only good way i know around it.

And back on topic:
"Luminosity" in Photoshop is way better at matching correct brightness then the Lightness channel in LAB. I pointed this out in my previous posts with a couple of clear visual examples and this is still my point.

Ps: Please do not say "Luminosity" when referring to the LAB Lightness-channel. It might confuse people.
Reply With Quote top
  #27  
Old 09-09-2010, 07:59 AM
mistermonday's Avatar
mistermonday mistermonday is offline
Moderator
Patron
 
Join Date: Dec 2005
Posts: 3,028
Re: L*a*b* question: why can't it be simulated in

Chain, my last post comments were based on my image, not your modified version, which I just found. Your modified version just proves my point and perhaps I did not explain the goal clearly enough.

Quote:
I'm assuming that the purpose of your "dodging" was to increase apparent brightness without affecting color. In that case Luminosity blending mode in RGB triumphs over the L-channel in LAB.
Not exactly. Like RGB, a dodge or burn curve can NOT alter the lightness without affecting the color / saturation. LAB has the same problem but it varies the color / saturation somewhat differently. The net effect for both LAB and RGB is that after you dodge an area in a real image to brighten a frckle or wrinkle or darker area of skin, and lets say you dodge it to exactly match the lightness of the adjacent area, the color / sat shift will mean that the dodged area does not match its surroundings and you almost always need to perform a correction, usually by painting on a color layer or adding a Hue/Sat adj layer or both.

In LAB, this is not the case. Once you have matched two areas in lightness, you can paint on a color layer with an adjacent color and there will be a perfect match. If you do that in RGB it will NOT. Moreover, if you change the Dodge curve blend to Luminosity, it will be close but it will NOT match.

That was the purpose of my last post, to walk you through it. I understand the difference in lightness and luminosity and the only reason I asked you to change your eyedroppers to L in the RGB image was to help you visualize and ensure that the brightness of the dodged area was the same as the left side of the RGB image with the desaturated / visualization layer is turned on.

Andrew Rodney asked for a real life example. In real life D&B you need a reliable, fast, and simple way to dodge or burn an area up or down to where it is within range of adjacent area (not always exact but sometimes). The fastest easiest method is to do it visually and that usually requires a desaturated view sometimes coupled with a contrasted view to accentuate the difference in brightness. After all the work on the D&B layers is done, you almost always require color / saturation correction. In LAB you can paint away and you have perfect match. In RGB you can not do the same as you can see in my sample image. Not even if you change the curve to luminosity blend. In RGB its more work and time to clean things up and the results rarely look as good.

If you need to see this done on real skin I can put something together, however if you have done any detailed skin work you should already have an appreciation if the issues. If you or Andrew have a better way of doing things in RGB, I would be pleased to see it. I am open. I have am not polarized, biased, or bigotted toward either color model. For both RGB and LAB I find things which are more intuitive, faster, more effective, that produce excellent results. In the end, there is a lot of "realworld" work that must be completed.
Regards, Murray
Reply With Quote top
  #28  
Old 09-09-2010, 09:16 AM
andrewrodney's Avatar
andrewrodney andrewrodney is offline
Senior Member
 
Join Date: May 2010
Location: Santa Fe
Posts: 1,090
Re: L*a*b* question: why can't it be simulated in

To digress a bit, in terms of dodge and burn (and then some), the technique I’ve used, taught to me by JP Caponigro and modified by Mac Holbert is:

Make a new layer, name it Dodge & Burn, set the mode to Overlay. If you then paint on it with black you burn, paint with white, you dodge. You control opacity with the keyboard. Eraser removes what you didn’t want applied. JP paints a bit hot, then tweaks using the Fade Command which does a nice job of allowing control over that last brush stroke (problem is, its just the one stroke*).

JP also paints in warmness or coolness by setting the colors either to warm or cool (blue or yellow), using a new layer with the blend mode set to Hue. On that new layer, the brush paints warm/cool, not that actual color. To increase or decrease saturation, one more layer, blend mode set to Saturation. Pick a very saturated or muted color (the actual color itself doesn’t matter). Thanks to the blend mode, the brush only adds or reduces saturation.

*I find Lightroom’s selective brushes to work really well because you can paint in a dodge or burn, adjust at any time the strength much like the Fade command. I paint very hot so I can see exactly where I’m brushing the dodge or burn, (or any other edit using brushes) then I can move the appropriate slider to get exactly the density I want. Even weeks later since the beauty of a parametric editor is, history is always accessible and the edit never burns an edit into pixels until you render the data. I wouldn’t do this for tiny areas of an image. That’s a job for a pixel editor. Lots of brush strokes can slow LR down a great deal.But for decent sized areas, dodging and burning in Lightroom or ACR work really well for me and provide all the controls I need, no issue with color, layers making the file larger etc. I also like rendering out as much of the “corrections” as possible at the raw stage.
Reply With Quote top
  #29  
Old 09-09-2010, 09:54 AM
Chain's Avatar
Chain Chain is offline
Senior Member
 
Join Date: May 2009
Location: Oslo, Norway
Posts: 551
Re: L*a*b* question: why can't it be simulated in

It would be nice if you could put up some examples together using problematic images that we can use to compare the methods discussed here.

I just tried dodging some skin in LAB and RGB and comparing. Both color spaces were equally easy for this task, and both provided good results.
I made it nice and tidy, and my files are here (I couldn't get it all in the silly 100 KB retouchpro limit).
Additional comments are inside the files as notes.

My conclusion after the test is that in this example, there was absolutely no need flatten and convert to LAB just for the dodging.

Note: Personally i would probably also have attacked this with the healing brush to clean up the remaining texture, and maybe even a frequency separation, but I focused only on the dodging.
Reply With Quote top
  #30  
Old 09-09-2010, 11:39 AM
Flashtones Flashtones is offline
Senior Member
 
Join Date: Apr 2009
Posts: 956
Re: L*a*b* question: why can't it be simulated in

Great thread guys, keep it up.

mistermonday, I'm curious about how typically you do go into LAB for beauty D&B and your workflow to do it.
Reply With Quote top
Reply

  RetouchPRO > Technique > Photo Retouching


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Converting CMYK to RGB and back to CMYK. xxxmen Input/Output/Workflow 23 01-06-2011 11:56 AM
my 1st post..red cast on Lacie 324/CS4 RGB to CMYK nadaman Hardware 47 03-10-2010 02:31 PM
Adobe RGB (1998) profile problem * Sinisa Software 7 07-10-2009 11:08 PM
RGB Help PSE JohnTravers Salon 5 10-13-2008 07:28 PM
silly question Tinkerbella197 Photo Restoration 5 09-15-2008 11:46 AM


All times are GMT -6. The time now is 05:48 PM.


Powered by vBulletin® Version 3.8.4
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.
SEO by vBSEO ©2011, Crawlability, Inc.
Copyright © 2016 Doug Nelson. All Rights Reserved