Ad
Ad
Ad
Pages: « 1 ... 3 4 [5]   Bottom of Page
Print
Author Topic: Increasing color saturation in LAB: S-curve or straight line?  (Read 13904 times)
bjanes
Sr. Member
****
Offline Offline

Posts: 2795



« Reply #80 on: May 01, 2013, 12:42:47 PM »
ReplyReply

It sure could if you're silly enough to do this in 8-bit per color!

A keeper on the sillyness of conversions not needed by the late great Bruce Fraser in response to the Lab conversion craze started years ago. What Bruce wrote then is as true  today. Plus I don't see anything in your link that couldn’t be achieved faster, with no data loss in a good raw converter.

"One thing is certain. When you work in tagged high-bit RGB, you know that
you're working with all the data your capture device could produce. When you
work in Lab, you know that you've already discarded some of that data."



If I recall correctly, Bruce also once pointed out that whenever you edit an image, data are thrown away as pixels that were originally at different levels are compressed into one level. This creates spikes and holes in the histogram. However, we choose to make the edit anyway, as it improves the image. As you imply, working with 16 bit data minimizes the holes and they are usually visually loss-less. With 8 bit data, posterization may occur with edits, but this is much less likely when one is dealing with a 16 bit space. IMHO, too much is being made of this data loss, which is visually loss-less in a 16 bit space. There may be good reasons to avoid trips to and from L*a*b, but data loss is not one of them in most cases.

Bill
Logged
Gulag
Full Member
***
Offline Offline

Posts: 216


« Reply #81 on: May 01, 2013, 01:03:30 PM »
ReplyReply

For one, to separate aesthetics of images from the technical realities of image processing in terms of workflow, best image quality, flexibility and so forth. Now do you want to bring up great work which is an aesthetic discussion? Or continue on with the realities of image processing on data and it's effect?

For centuries, what really matters for any visual arts has been the final work. Do viewers really care if one has used 64-bit or 8-bit in his workflow? Or do they really care whether or not one has made some trips back and forth between RGB, LAB and CYMK color modes? I hate to sound cynical but I was led to believe great art comes from artists' internal worldview and their creativity/innovation. In the age of digital tools, that's not true anymore?
« Last Edit: May 01, 2013, 01:07:12 PM by Gulag » Logged

“For art to be art it has to cure.”  - Alejandro Jodorowsky
digitaldog
Sr. Member
****
Offline Offline

Posts: 9170



WWW
« Reply #82 on: May 01, 2013, 01:05:48 PM »
ReplyReply

For centuries, what really matters for any visual arts has been the final work. Do viewers really care if one has used 64-bit or 8-bit in his workflow? I hate to sound cynical but I was led to believe great art comes from artists' internal worldview. In the age of digital tools, that's not true anymore?

It is true and totally immaterial to this discussion about increasing color saturation and the use of Lab.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
digitaldog
Sr. Member
****
Offline Offline

Posts: 9170



WWW
« Reply #83 on: May 01, 2013, 01:11:52 PM »
ReplyReply

With 8 bit data, posterization may occur with edits, but this is much less likely when one is dealing with a 16 bit space. IMHO, too much is being made of this data loss, which is visually loss-less in a 16 bit space.

Agreed. So we have high bit data. Where did it come from? Could the original input controls have provided the color appearance desired or did someone have to use Photoshop after the fact to get the look they wanted?
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1320


« Reply #84 on: May 01, 2013, 05:21:16 PM »
ReplyReply

Hi,

Topaz Labs "Adjust" could be what you are looking for. It's a Photoshop plug-in, but can also be used without Photoshop in Topaz Labs "photoFXlab", which is a kind of command central for all plugins, but it also offers e.g. layers and masking functionality. 'Adjust' offers, amongst others, Adaptive Saturation, regular saturation, and low Saturation Boost controls, while attempting to leave Luminosity alone.

Cheers,
Bart

I 'd suggest (not sure if that already suggested up the thread) CurveMeister ( http://www.curvemeister.com/ ) , that is the proper tool.
Logged
G*
Newbie
*
Offline Offline

Posts: 48


« Reply #85 on: May 02, 2013, 02:53:53 AM »
ReplyReply

@ Mark D Segal
Totally agree about the reversibility issue. But my understanding is that sooner or later Photoshop sends every image file through CIELab, even with RGB>RGB conversions, so you could apply LAB-adjustments to the file (not the preview) at that later stage, for example together with conversion to output color space.

@ digitaldog
CIELab probably isn’t even close to perceptually uniform, but that CIELab_to_UPLab.icc by Bruce Lindbloom doesn’t seem to be such a bad workaround. I think RPP is using this file, and to good success. I don’t think adjustments in RGB don’t work at all, and maybe its construction has a lot of advantages, but I am a little more comfortable with a clear separation of L* and a*/b* curves. Furthermore blending modes in RGB do not seem to work to the same effect (see also posts by Schewe and jrp) – which is funny since to my understanding under the hood Photoshop is somehow going back to something LAB-ish anyway. I also like the effect of color separation.

@ BartvanderWolf
Thanks for suggesting Topaz Adjust. I will have a look at the trial version.

@ tlooknbill
I am more of a Capture One guy and never got used to LR/ACR. Nothing based on facts, just personal liking.

@ JeffKohn
Thanks for your posts.
"and then adjust the blend-if sliders of the layer properties to feather out the effect on the more saturated colors"
Didn’t try that before but works like a charm. Again thanks a lot!

@ Vladimirovich
Thanks for the link. I will have a look at that tool. (And I really don’t care whether it’s connected to "DM" or not since I’m too young to remember when those LAB/RGB wars took place).
« Last Edit: May 02, 2013, 03:20:45 AM by G* » Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9170



WWW
« Reply #86 on: May 02, 2013, 08:53:22 AM »
ReplyReply

But my understanding is that sooner or later Photoshop sends every image file through CIELab, even with RGB>RGB conversions, so you could apply LAB-adjustments to the file (not the preview) at that later stage, for example together with conversion to output color space.

Not so! An old urban legend. Not necessary, time consuming. Photoshop does NOT convert to Lab to make such conversions, it uses Lab as a profile connection space to convert between color spaces. What happens is that when you ask for a color conversion, Photoshop builds a conversion table from source to destination using LAB to find the equivalent values. It's done in 20-bit precision resulting in fewer quantization errors than you would actually converting the pixels to Lab!  
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
G*
Newbie
*
Offline Offline

Posts: 48


« Reply #87 on: May 02, 2013, 09:59:10 AM »
ReplyReply

Aha. Thanks. So I learned something again.

That’s probably good news for RGB>RGB conversions – but makes me wonder why not the same precision is used when converting RGB>LAB/LAB>RGB.
Logged
MarkM
Sr. Member
****
Offline Offline

Posts: 328



WWW
« Reply #88 on: May 02, 2013, 11:35:16 AM »
ReplyReply

@Andrew:

For some reason I thought the icc spec mandated XYZ to be used as the profile connection space with matrix profiles (I haven't tried to look it up again). This would seem to be computationally easier than using LAB. I'm sure it's inconsequential, but I'm curious—do you know for sure that LAB is used as PCS for everything under the hood in photoshop?
Logged

Jack Hogan
Full Member
***
Offline Offline

Posts: 243


« Reply #89 on: May 02, 2013, 11:38:46 AM »
ReplyReply

I believe Lab is used whenever a 'perceptual' intent is called and the profile allows it.
Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5489


WWW
« Reply #90 on: May 02, 2013, 12:24:45 PM »
ReplyReply

For some reason I thought the icc spec mandated XYZ to be used as the profile connection space with matrix profiles (I haven't tried to look it up again).

As far as I know, v2 color space (display class) has PCS in XYZ, but v4 profiles can be in Lab. The sRGB v4 ICC preference beta has a PCS of Lab.
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2795



« Reply #91 on: May 02, 2013, 07:12:16 PM »
ReplyReply

For some reason I thought the icc spec mandated XYZ to be used as the profile connection space with matrix profiles (I haven't tried to look it up again). This would seem to be computationally easier than using LAB. I'm sure it's inconsequential, but I'm curious—do you know for sure that LAB is used as PCS for everything under the hood in photoshop?

One can easily check the PCS for a given profile with the freeware Profile Insepctor from the ICC.

AdobeRGB is a matrix profile and the PCS is XYZ. The profile for my Epson 3880 shown in the illustration is a lookup type profile as indicated by the large size (compare to the AdobeRGB size) and it is LAB. This information might be of interest to some, but I don't know what to make of it.

Bill

Logged
Schewe
Sr. Member
****
Offline Offline

Posts: 5489


WWW
« Reply #92 on: May 02, 2013, 08:12:27 PM »
ReplyReply

AdobeRGB is a matrix profile and the PCS is XYZ. The profile for my Epson 3880 shown in the illustration is a lookup type profile as indicated by the large size (compare to the AdobeRGB size) and it is LAB. This information might be of interest to some, but I don't know what to make of it.

Input and output class profiles generally use Lab and are table based profiles. Display class (as in working spaces) are generally matrix based, not table based and generally use XYZ although since v4 display profiles can have multiple rendering intends, the PCS can be Lab.
Logged
G*
Newbie
*
Offline Offline

Posts: 48


« Reply #93 on: May 07, 2013, 07:56:52 AM »
ReplyReply

Something promising I found a little after the death of this thread:

Vibrance correction in RawTherapee (currently 4.0.10.72 for Mac).
Looks promising, sounds well done.

You migth want to have a look at this for further details:
http://jacques.desmis.perso.neuf.fr/RT/vibrance2.html

Only pity: RT 4.0.10.72 runs on my Mac at work (Mac OS 10.6.8.) but not at home (Mac OS 10.8…). Will have to figure out where the fault is …
« Last Edit: May 07, 2013, 08:00:58 AM by G* » Logged
kirkt
Full Member
***
Offline Offline

Posts: 180


« Reply #94 on: May 07, 2013, 12:34:19 PM »
ReplyReply


Only pity: RT 4.0.10.72 runs on my Mac at work (Mac OS 10.6.8.) but not at home (Mac OS 10.8…). Will have to figure out where the fault is …

The RT blog/download has compilations for 10.7 and 10.8 (i run 10.7 and the current 10.7 compilation works for me).  One thing I have run up against previously that has caused older compilations to crash during launch has been the pesky "libcups.2.dylib" issue.  The OS library seems to be in conflict with the one contained within the RawTherapee.app.  The workaround that I have used with success is to copy libcups.2.dylib from my OS and paste a copy into the RawTherapee.app/Contents/MacOS/lib folder of the Raw Therapee application (replace the RT version with the OS version).  This will get it to launch and run.  Try the OS appropriate (10.8 ) version on your newer OS machine and see if it works before attempting this kludge.

I noticed in the most recent OS X 10.7 application that the libcups.2.dylib library is not present in the lib folder within the application, so maybe that clears up some of the problems on 10.7 and >.

kirk
« Last Edit: May 07, 2013, 12:36:04 PM by kirkt » Logged
G*
Newbie
*
Offline Offline

Posts: 48


« Reply #95 on: May 07, 2013, 01:33:25 PM »
ReplyReply

Thanks, Kirk, I’ll have a look at that.

In my case the app launches but does not output converted tifs. I see that the process starts but it does not come to an ending. First I thought that my machine might well be a bit lame (1.8 GHz Intel i7 with 4 GB DDR3 @ 1333MHz), but after an hour or so and nothing further happening … No error message though.
Logged
kirkt
Full Member
***
Offline Offline

Posts: 180


« Reply #96 on: May 07, 2013, 01:37:17 PM »
ReplyReply

Thanks, Kirk, I’ll have a look at that.

In my case the app launches but does not output converted tifs. I see that the process starts but it does not come to an ending. First I thought that my machine might well be a bit lame (1.8 GHz Intel i7 with 4 GB DDR3 @ 1333MHz), but after an hour or so and nothing further happening … No error message though.

Oh.  That's a horse of a different color.

good luck,

kirk
Logged
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1320


« Reply #97 on: May 08, 2013, 07:19:01 PM »
ReplyReply

I think RPP is using this file, and to good success.
RPP does certain postdemosaick (note that WB, curves, expocorrections, etc - are predemosaick in RPP) adjustments like saturation in UPLab, yes...
Logged
Pages: « 1 ... 3 4 [5]   Top of Page
Print
Jump to:  

Ad
Ad
Ad