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
  #11  
Old 09-07-2010, 08:15 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

Andrew, OK while tying to make a point I was over the top with the 10,000 times comment. For the purposes of how many times you would actually make practical conversions back and forth with LAB, you visibly will not see destruction. To some extent the point is moot because often the reason you go into LAB are to make significant changes to the image that are intended to enhance it. If the results are better than the original, the the damage done to the image was positive.
Having put LAB to a lot of practical application over the past few years, I will repectfully agree to disagree with you on this topic. And by the way, it is so far the only one of your views expressed in these forums that I disagree with.
Regards, Murray
Reply With Quote top
  #12  
Old 09-07-2010, 08:28 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 mistermonday View Post
Andrew, OK while tying to make a point I was over the top with the 10,000 times comment.
Just a bit when the facts are, the image data loss is produced almost entirely on the first conversion.

Quote:
For the purposes of how many times you would actually make practical conversions back and forth with LAB, you visibly will not see destruction.
Visually see where, when and on what device?

Quote:
If the results are better than the original, the the damage done to the image was positive.
Only if you didn’t enhance it earlier in the workflow when you should (when you scan it, or when you render the raw data). Or by avoiding the issue at capture. Forgive me, I’m a classical trained photographer. I try to nail it in capture, not like Dan, after the fact to demonstrate how well I can polish a turd.

Of course. That’s why Photoshop exists. Its sole job is to alter existing RGB values. but I’m a huge believer in GIGO:GARBAGE IN GARBAGE OUT!

Its funny why those that spend so much time teaching you how to fix an awful original pay no attention to not getting an awful original in the first place. But again, my job for years was to produce good originals. Or to scan good data (not necessarily to match an ugly original unless asked to do so). Or today, to capture and render the raw data as cleanly as possible before Photoshop ever enters the picture.
Reply With Quote top
  #13  
Old 09-07-2010, 08:41 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

Andrew, somewhere along the line I seem to have conveyed the wrong message. Being a photographer first and foremost, with an embarassingly long number of years behind the lense, I can appreciate as much the importance of getting it right on the shoot. There are a great many applications of LAB not directed at fixing poor photographs. It was largely those I was referring to. Although LAB can be an excellent assist in retouching as well. I was not selling LAB as a miracle cure for poor photography. I agree with GIGO and the concept that you can polish poop.
Regards, Murray
Reply With Quote top
  #14  
Old 09-08-2010, 10:06 AM
DJSoulglo's Avatar
DJSoulglo DJSoulglo is offline
Senior Member
 
Join Date: Jan 2006
Location: Amsterdam
Posts: 475
Re: L*a*b* question: why can't it be simulated in

I love LAB as a tool to use. My main editing mode is always in RGB (although I started out in CMYK), I don't think it's that destructive, but I only use it for immense color shifts (anything to black/white).

Also a curves layer in luminosity and doing that same curves adjustment in LAB isn't quite the same thing. The RGB version isn't a perfect translation of that process. It's close, but no cigar.
Reply With Quote top
  #15  
Old 09-08-2010, 12:49 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 DJSoulglo View Post
Also a curves layer in luminosity and doing that same curves adjustment in LAB isn't quite the same thing. The RGB version isn't a perfect translation of that process. It's close, but no cigar.
I would expect the math to be different. If you tried both approaches and using Calculations you subtracted the two, I would expect we would see some differences. The question I have is, how would one test the two and examine that the RGB version isn’t a cigar? What test files and steps would one follow and what would they see that would show that the Lab conversion was superior?

A question to Murry. Where are you finding the conversion to Lab is necessary (what problem and technique are you finding warrants a Lab conversion)?
Reply With Quote top
  #16  
Old 09-08-2010, 01:42 PM
Chain's Avatar
Chain Chain is offline
Senior Member
 
Join Date: May 2009
Location: Oslo, Norway
Posts: 551
Post Re: L*a*b* question: why can't it be simulated in

I might as well toss some more fuel on the fire.

Attached is an image showing simply what happens to your image data if you convert an 8-bit image to LAB in Photoshop (dithering). Additional conversions makes it worse. See image for more info.
Ps: I never said that converting to LAB was "very destructive", and in 16-bit any loss would be negligible (but would it really be worth the hassle?).
Edit: I should have seen this, but by not going through Image > Mode but instead going through Edit > Convert to Profile... you can turn off the dithering. This way you can avoid the dithering shown in the example (trade-off is more visible banding in some situations). A bit cumbersome though.

To be honest, the definition of "Luminosity" in Photoshop is way closer to human perceived brightness than the Lightness channel of a LAB image. So I do not see why LAB would be better for adjusting perceived brightness and color separately...
Photoshop uses a weighted sum of R, G and B to calculate the "Luminosity". The weighting choice is an attempt to compensate for the sensitivity of the human eye to each color.
Example attached (although it is best to overlay the images in Photoshop and turn them on/off to see how the brightness shifts).
Just to wrap this post up with another interesting piece of information; I would like to point out how simply desaturating the image in Photoshop does not use any such weighting of the colors and thus results in a change in brightness. In my example image the blue sky would end up brighter (our eyes are not that sensitive to blue) and the grass would turn too dark (we are quite sensitive to green).
Note; in this post when i say "brightness" i mean how bright the image appear to the human eye.
Attached Images
File Type: png LAB-test.png (30.3 KB, 67 views)
File Type: jpg luminosity.jpg (97.7 KB, 76 views)

Last edited by Chain; 09-08-2010 at 04:02 PM. Reason: Solved the unwanted dithering issue...
Reply With Quote top
  #17  
Old 09-08-2010, 01:52 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

Excellent visual presentation, it sure does appear that the method Photoshop is using produces, to my eye, a more “pleasing” rendering!

Quote:
To be honest, the definition of "Luminosity" in Photoshop is way closer to human perceived brightness than the Lightness channel of a LAB image.
Quote:
Note; in this post when i say "brightness" i mean how bright the image appear to the human eye.
Also nice to see Luminosity (incorrectly labelled in Photoshop since it was implemented) and Brightness (what we perceive) differentiated here. My partner, a true color scientists would be SO happy to see this post, as the term Luminosity as defined in Photoshop is a major pet peeve of his.

Gang, there’s so very useful info in that last post that anyone using Photoshop should study.
Reply With Quote top
  #18  
Old 09-08-2010, 02:20 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

Andrew, here is one reason - Color correcting Dodge and Burn Curve Adj layers. I use the two sample files attached. Please excuse the simplicity - solid colors, no texture, to illustrate the point, also an exaggerated color difference to illustrate the reason.
Please open the RBG file with just the bg layer active. Add a dodge curve, dodge the top right corner until the luminosity of it equals the left side iof the image (eyedropper samplers are there). Add a blank layer set to color and paint with the left side color. Turn off the visibility layer.
In a typical D&B on a portrait, the D/B curves shift the saturation as well as the lightness. Trying to color correct is not that easy or quick. You can add a Hue/Sat layer with a layer mask copied from the D/B curve but that has its issues.
In LAB, 2nd file, it doesn't matter where the A and B values go when you move the curves. The luminosity can be match very quickly. Add a blank color layer and you can paint in any color and get a perfect match because the L channel won't move.
I am NOT saying that this is a necessary part of the workflow. It just happens to add speed, works smoothly, and intuitively. Moreover I find more lattitude in the base brightening and lightening curves which can have an impact on how many times you need to brush over a pixel to brighten or lighten it.
Regards, Murray
Attached Files
File Type: zip MM Dodge Test RGB.zip (39.0 KB, 22 views)
File Type: zip MM Dodge Test LAB.zip (180.2 KB, 17 views)

Last edited by mistermonday; 09-08-2010 at 02:39 PM.
Reply With Quote top
  #19  
Old 09-08-2010, 02:22 PM
Chain's Avatar
Chain Chain is offline
Senior Member
 
Join Date: May 2009
Location: Oslo, Norway
Posts: 551
Post Re: L*a*b* question: why can't it be simulated in

(my above post is the "important" one)

I've attached a more extreme example here to show that "Luminosity" in Photoshop is way better than LAB's Lightness channel at representing the apparent brightness of the image, and thus better at separating brightness/color when doing adjustments.

Note on the related blending modes:

In Photoshop the "Color" blend mode is the inverse of "Luminosity" (it affects only the color, but not perceived brightness).
The "Hue" and "Saturation" blending modes are also based on the same calculation but further divide "color" into the two components "Hue" and "Saturation".
Notice that "Saturation" blending thus works different (more visually accurate) than the desaturation (or hue/saturation) adjustments.
To see this yourself add a hue/saturation adjustment layer, set saturation to -100 (desaturate). Then change blending mode to Saturation and see how this restores the original brightness.
This is why you always want to change your hue/saturation-adjustment's blending mode (also for hue adjustments).
Attached Images
File Type: jpg maintaining-brightness.jpg (65.0 KB, 37 views)
Reply With Quote top
  #20  
Old 09-08-2010, 02:29 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

Quote:
Originally Posted by mistermonday View Post
Please open the RBG file with just the bg layer active. Add a dodge curve, dodge the top right corner until the luminosity of it equals the left side iof the image (eyedropper samplers are there). Add a blank layer set to color and paint with the left side color. Turn off the visibility layer.
I will look closer at the files you supplied now but a quick look tells me that the "Visualize" layer on the RGB file should probably have been set to blending mode Color or Saturation to give a correct result?
And if you want the RGB version to not shift the color the curves (dodge) layer should be set to Luminosity blending.
Edit: Also the RGB image is a bit blurry in the transition between the left and right side (it has been resampled i think).
Edit: The hue/saturation layer might be problematic for visualisation in LAB as the blend modes do not work... I'm a bit unsure here.

Last edited by Chain; 09-08-2010 at 02:42 PM.
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 11:38 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