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.

Frequency Seps and profile conversions

Reply
 
Thread Tools
  #1  
Old 06-28-2010, 09:43 PM
Flashtones Flashtones is offline
Senior Member
 
Join Date: Apr 2009
Posts: 956
Frequency Seps and profile conversions

I have a file with a couple of frequency separations in it that is changing dramatically (lightening) when I try to do a color profile conversion. The frequency separated layers seem to be altered more than others so adjustments above the separations no longer match in tone to the separations.

All told, converting the layered file results in an unusable mess.

Is there a reason why high pass (apply image) and/or linear light layers don't convert properly?
Reply With Quote top
  #2  
Old 06-28-2010, 10:42 PM
mistermonday's Avatar
mistermonday mistermonday is offline
Moderator
Patron
 
Join Date: Dec 2005
Posts: 3,028
Re: Frequency Seps and profile conversions

Hi Flashtones. Remember that old rule about the whole not being equal to the sum of its parts? Well it applies here. You will also notice that you will get different results when you select a different Intent (Absolute, Relative Colorimetric, Perceptual, Saturation).
When you plan to build a complex stack of adjustment layers, frequency splits, etc, you need to choose a color space and stick with it for the master. Only make profile conversions to flattened copies.
The issue is that when you switch color spaces you are mapping pixel RGB values from one space to the other and are therefore changing those values. Some layers will have a distribution of pixels that will be slightly shifted but other layers will undergo significant shifts. When you have layers whose values are very interdependent in the source color space and they become divergent in the destination space, the relationship changes and you can see profound visual differences.
Frequency Split layers are particularly sensitive to some profile changes. As you know when you split the image GB+HP=Orig. Any change to either layer disturbs the precise relationship. As you know much of the split and Linear Light Blend relies on the 50% gray of the majority of pixels on the HF layer. In fact all of the Contrast Blend Modes (Overlay, Soft Light, Linear Light, etc) have that magic 50% gray point where the pixels are not contrasted / changed. So as an example, if you create a 50% gray layer in sRGB and convert the profile to Color Match with Absolute Colorimetric as your intent, the RGB values go from 128, 128, 128 to 109, 109, 109. Now your split layer is no longer a properly split layer. The formula is badly broken. That's just one example of others in the layer stack that become altered in conversion. The result is that all your retouching work is rendered pretty well useless when you profile convert a layered stack.
Regards, Murray

Last edited by mistermonday; 06-28-2010 at 10:54 PM.
Reply With Quote top
  #3  
Old 06-28-2010, 11:01 PM
Flashtones Flashtones is offline
Senior Member
 
Join Date: Apr 2009
Posts: 956
Re: Frequency Seps and profile conversions

Thanks for the explanation, Murray.

I see I need to put more forethought into my color space choice. Was messing around with prophoto, but my files are getting so large that at a certain stage I need to downsample to 8-bit to save on overhead. My 2 yo MacBook Pro w/ 4 gigs RAM can't keep pace on multi-gig files. Figured it best to do any colorspace conversions first, while still in 16-bit, to avoid any rounding errors in the conversion.

At this point I'll stay in prophoto but downsample to 8-bit, then convert a flattened version at the end. Unfortunately, prophoto would not be my preferred 8-bit working space.
Reply With Quote top
  #4  
Old 06-29-2010, 10:37 AM
Flashtones Flashtones is offline
Senior Member
 
Join Date: Apr 2009
Posts: 956
Re: Frequency Seps and profile conversions

FWIW, I was able to come to a satisfactory solution by assigning the profile before converting to it. This gave me a color shift (that in this instance was not objectionable) but left tones relatively intact.

I'll never really understand color management....
Reply With Quote top
  #5  
Old 06-29-2010, 11:36 AM
mistermonday's Avatar
mistermonday mistermonday is offline
Moderator
Patron
 
Join Date: Dec 2005
Posts: 3,028
Re: Frequency Seps and profile conversions

That is probably the best compromise if you have already done all the retouching in one color space and need to move the entire layer stack to another. You have basically applied a Jedi Mind Trick. By assigning a different profile you have mainyained all of the RGB values but they will now be displayed in the manner that they would be viewed in that new color space. One analogy would be: you have taken an image of a gray squirrel and told photoshop that you have given it an image of a black squirrel.Consequently Photoshop will now display an image of a black squirrel. That analogy is a little extreme because when it comes to color spaces, the differences are more subtle, oftening appearing as changes in saturation, contrast, and some color changes. Sometimes assigning one color space to another actually improves the image, but in those cases reversing the color spaces looks awful. You can see this if you assign Adobe RGB to an image which was edited in sRGB and vice verse. The different directions produce very opposite results.
But you are coorect; profile re-assignment is much less destructive than profile conversion for complex layer stacks.
Regards, Murray
Reply With Quote top
  #6  
Old 06-29-2010, 01:02 PM
andrewrodney's Avatar
andrewrodney andrewrodney is offline
Senior Member
 
Join Date: May 2010
Location: Santa Fe
Posts: 1,091
Re: Frequency Seps and profile conversions

Well initially, assigning a profile appears to be less destructive because the numbers don’t change. But at some point, unless you are just viewing the image on screen, you do have to convert that data, at this point the free lunch is over with.
Reply With Quote top
  #7  
Old 06-29-2010, 01:42 PM
Flashtones Flashtones is offline
Senior Member
 
Join Date: Apr 2009
Posts: 956
Re: Frequency Seps and profile conversions

Quote:
Originally Posted by andrewrodney View Post
Well initially, assigning a profile appears to be less destructive because the numbers don’t change. But at some point, unless you are just viewing the image on screen, you do have to convert that data, at this point the free lunch is over with.

You make it sound equally damaging. But one way yields a very usable result, the other doesn't, at least in the case of this file.
Reply With Quote top
  #8  
Old 06-29-2010, 01:47 PM
andrewrodney's Avatar
andrewrodney andrewrodney is offline
Senior Member
 
Join Date: May 2010
Location: Santa Fe
Posts: 1,091
Re: Frequency Seps and profile conversions

Quote:
Originally Posted by Flashtones View Post
You make it sound equally damaging. But one way yields a very usable result, the other doesn't, at least in the case of this file.
What I’m saying is, at some point, the numbers have to change and at that point, the conversion takes place with that assigned profile and there is data loss. Do it in 16-bit, its moot. But assigning an ICC profile may appear to change the color appearance without the numbers (it is) indicating that the original profile assignment was incorrect and, at some point, numbers do update and change. Just like an Adjustment Layer. If you print the document, the numbers get updated. Its only non destructive as you view that composite of layers on-screen. Flatten or print, the numbers change and the free lunch as I indicated, is now over.
Reply With Quote top
  #9  
Old 06-29-2010, 01:53 PM
Flashtones Flashtones is offline
Senior Member
 
Join Date: Apr 2009
Posts: 956
Re: Frequency Seps and profile conversions

By "free lunch" I' assume you mean a spiked histogram which is probably not detectable by eye in the image itself, but which could rear it's ugly head if further significant editing is done.

What I am telling you is that directly converting this file from ProPhoto to aRGB or sRGB yielded a radically unusable image upon conversion.

Forget lunch, we're talking caca.
Reply With Quote top
  #10  
Old 06-29-2010, 01:57 PM
andrewrodney's Avatar
andrewrodney andrewrodney is offline
Senior Member
 
Join Date: May 2010
Location: Santa Fe
Posts: 1,091
Re: Frequency Seps and profile conversions

Quote:
Originally Posted by Flashtones View Post
By "free lunch" I' assume you mean a spiked histogram which is probably not detectable by eye in the image itself, but which could rear it's ugly head if further significant editing is done.
No, not necessarily. That’s an extreme case. Anytime you alter pixel values, there are rounding errors and thus data loss. The cases above fall into this camp. This can be compounded by many such edits until as you point out, the histogram shows this and eventually, it shows upon output.

The real question should be, why does one need to assign the profile to this set of numbers? Why is the original tag not the proper tag?
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



All times are GMT -6. The time now is 01:13 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