Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: The purple blues  (Read 4634 times)
purpleblues
Newbie
*
Offline Offline

Posts: 20


« on: July 12, 2009, 05:53:56 PM »
ReplyReply

Images shot with my dusty but trusty EOS 5D show a nasty purple tint in bright blue areas. I tried five different raw converters on this picture.

The shot was taken at ISO 1600 during a conference with several presentations. The stage was lit by stage projectors with tungsten bulbs. The light effect on the wall left of the speaker was done with small tungsten projectors with blue filters on the stage floor.

White balance was set in a different image on a neutral white surface on the same stage and copied to this image. Apart from that no adjustments were made. The brightest blue-lit areas on the wall were of course clipped. But the hue shift occurs not in the clipped areas but outside of them. Highlight recovery or reducing the exposure showed undesirable results since I had to optimize for the best appearance of the speaker. And the effect could be reduced but not avoided by varying exposure or recovery.

[attachment=15346:comparison.jpg]

The first row was done with DPP (neutral), C1 4.8.1 and a trial version of Aperture 2.1.1, the second row with a trial version of LR 2.4 "Adobe standard" and with a calibrated DNG profile, the last one with RD 1.8.3. In the "handmade" DNG profile I tweaked the colors so that there was no shift in brightness or hue in the colors of the color checker but no further tweakings were done.

The renderings of C1 and RD come very close to what I remember having seen with my own eyes and handle the clipped and almost clipped areas without significant hue shift and smooth gradients. All the other renderings show arguably beautiful but largely false colors and therefore no smooth gradients.

Did anyone experience similar results maybe even with other cameras (other Canon bodies, Nikon, MF backs) and how did you deal with it?

So far I see a reasonable advance for the small companies (Phase One and Iridientdigital). Their color profiles deliver very good results out of the box without any need for individual calibration (at least for my purposes) that are superior to the camera manufacturer's software, apple's pro software and LR. Too bad since i like the interface, workflow and the intuitive adjustment possibilities of LR.
Logged
Damo77
Jr. Member
**
Offline Offline

Posts: 69


WWW
« Reply #1 on: July 12, 2009, 11:55:46 PM »
ReplyReply

Um ... are you sure this isn't a problem with your monitor?
Logged

Damien
purpleblues
Newbie
*
Offline Offline

Posts: 20


« Reply #2 on: July 13, 2009, 03:01:04 AM »
ReplyReply

Quote from: Damo77
Um ... are you sure this isn't a problem with your monitor?
Yes I am sure. The purple doesn't only show on my monitor (it's a hardware calibrated Quato 230 with a huge color space) but also on other monitors and in print.

Besides that, if it was only my monitor you wouldn't see it on your monitor. (The screenshot wasn't created with a camera but with "shift-cmd-3".)
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9317



WWW
« Reply #3 on: July 13, 2009, 07:38:06 AM »
ReplyReply

Quote from: purpleblues
So far I see a reasonable advance for the small companies (Phase One and Iridientdigital). Their color profiles deliver very good results out of the box without any need for individual calibration (at least for my purposes) that are superior to the camera manufacturer's software, apple's pro software and LR.

I agree but would not put all the credit/blame towards color profiles, that's a bit of a stretch.

Blues shifting purple is a very common problem with the basic limitations of the Lab Color model. Its likely some clever color scientists are pre-tweaking this somewhere in the pipeline. And this can be more pronounced depending on the display or printer technology handling the data.
Logged

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

Posts: 288


« Reply #4 on: July 13, 2009, 08:21:50 AM »
ReplyReply

I often see this purple / blue thing in images containing very saturated blues. ( Stained glass - R72 G7 B230 ). I also often notice a loss of detail in these blue areas. Would using a filter on the lens to cut down the amount of blue information hitting the sensor help? ( You guessed it - I know little or nothing about filters ).

In the attached image I got rid of the purple by using the Recovery slider. ( Canon 20D, 70 - 200f4, ACR4.6.0.30 ).

Graeme

[attachment=15368:purple_blues.jpg]
Logged
papa v2.0
Full Member
***
Offline Offline

Posts: 198


« Reply #5 on: July 13, 2009, 08:37:06 AM »
ReplyReply

MMM that interesting

What color space are you rendering to in each instance?


Logged
purpleblues
Newbie
*
Offline Offline

Posts: 20


« Reply #6 on: July 13, 2009, 10:56:33 AM »
ReplyReply

Quote from: digitaldog
Blues shifting purple is a very common problem with the basic limitations of the Lab Color model. Its likely some clever color scientists are pre-tweaking this somewhere in the pipeline.

Well, it's the result that counts so I'm glad at least some color wizards know how to deal with it. I wish Adobe knew that too.

I would like to know more about these basic limitations of the LAB color model you mentioned. Maybe the yellow shift in highly saturated reds has similiar reasons. That's a less obvious but nonetheless similarly persisting problem. Only the highly saturated greens seem to be free of such effects. Do you know where I can read more on this or can you tell me a bit more about it yourself?

Thanks!
Purpleblues
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9317



WWW
« Reply #7 on: July 13, 2009, 11:31:21 AM »
ReplyReply

Quote from: purpleblues
I would like to know more about these basic limitations of the LAB color model you mentioned.

Keep in mind that CIELab was just an attempt to create a perceptually uniform color space where equal steps correlated to equal color closeness based on the perception of a viewer. The CIE didn't claim it was prefect (cause its not). Most color scientists will point out that Lab exaggerates the distance in yellows and consequently underestimate the distances in blues. Lab assumes that hue and chroma can be treated separately. There's an issue where hue lines bend with increase in saturation perceived by viewers as an increase in both saturation and a change in hue when that's really not supposed to be happening. Further, according to Karl Lang, there is a bug in the definition of the Lab color space. If you are dealing with a very saturated blue that's outside the gamut of say a printer, when one uses a perceptual rendering intent, the CMM preserves the hue angle and reduces the saturation in an attempt to make a less saturated blue within this gamut. The result is mathematically the same hue as the original, but the results end up appearing purple to the viewer. This is unfortunately accentuated with blues, causing a shift towards magenta. Keep in mind that the Lab color model was invented way back in 1976, long before anyone had thoughts about digital color management.
Logged

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

Posts: 288


« Reply #8 on: July 13, 2009, 12:19:49 PM »
ReplyReply

Quote from: papa v2.0
MMM that interesting

What color space are you rendering to in each instance?
I work in ProPhoto RGB. I converted the attached image to sRGB to post on this forum but the purple effect is present in the ProPhoto files
« Last Edit: July 13, 2009, 12:48:50 PM by graeme » Logged
purpleblues
Newbie
*
Offline Offline

Posts: 20


« Reply #9 on: July 13, 2009, 03:09:41 PM »
ReplyReply

Quote from: digitaldog
There's an issue where hue lines bend with increase in saturation perceived by viewers as an increase in both saturation and a change in hue when that's really not supposed to be happening. Further, according to Karl Lang, there is a bug in the definition of the Lab color space. If you are dealing with a very saturated blue that's outside the gamut of say a printer, when one uses a perceptual rendering intent, the CMM preserves the hue angle and reduces the saturation in an attempt to make a less saturated blue within this gamut. The result is mathematically the same hue as the original, but the results end up appearing purple to the viewer. This is unfortunately accentuated with blues, causing a shift towards magenta. Keep in mind that the Lab color model was invented way back in 1976, long before anyone had thoughts about digital color management.
Thanks a lot digitaldog for this enlightening answer. I also noted the purple shift when converting RGB to CMYK and asked myself if no one except me saw this. It is somehow relieving to learn that the issue really exists. Obviously C1 and RD take it into account and others don't (yet?). Since digital color management and digital photography are state of the art now and will be in the forseeable period of time I think it's time to straighten these things out. Or do we really want to live with false colors when it's not necessary?
Logged
tho_mas
Sr. Member
****
Offline Offline

Posts: 1697


« Reply #10 on: July 13, 2009, 03:40:50 PM »
ReplyReply

Quote from: graeme
I work in ProPhoto RGB. I converted the attached image to sRGB to post on this forum but the purple effect is present in the ProPhoto files
Quote from: graeme
I often see this purple / blue thing in images containing very saturated blues. ( Stained glass - R72 G7 B230 ). I also often notice a loss of detail in these blue areas.
The ProPhotoRGB 72/7/230 is displayed as Lab 20⎢84⎢-125 in the color tool of Photoshop. Nice so far. Now convert your ProPhotoRGB 72/7/230 from ProPhotoRGB to Lab. One would expect that the Lab values remain the same. But they change to 38/55/-90... which is a violett.
ProPhoto: [attachment=15383:prophoto.jpg]
ProPhoto converted to Lab: [attachment=15384:lab.jpg]

Exactly the same happens if you just desaturate your ProPhotoRGB blue... it becomes violett.

High saturated blues and ProPhoto are difficult to handle as ProPhoto contains colour values that can be set mathematically but that are outside of real colours. The highest saturated blue in ProPhotoRGB (i.e. RGB 0/0/255) is Lab 0⎢90⎢-128. Apart from that this is a colour far beyond any colour a monitor or any other real world device is able to produce it's moreover absurdity because a Lab value without luminance and saturation is by definiton black.
Your ProPhotoRGB 72/7/230 is even outside of J. Holmes' Ekta space...

But: didn't you see any clipping in the histogram?

edit... forgot about the "loss of details" in blues...
Depends: maybe (in this case I'm almost certain) it's just that you monitor clips the colours. If you move with mouse over the respective areas and look at the colour values you will see if there is any modulation or not.
But at these values in any case you will lose all modulation at the latest when it comes to printing.

Quote from: purpleblues
Obviously C1 and RD take it into account and others don't (yet?).
C1 doesn't work with colour spaces from outer space. C1 is totally based on real world colour management - either way if there is a "bug" in Lab or not. We work with Lab (as PCS) ... so what!
If you like ProPhotoRGB for whatever reason you can process your RAW files to ProPhotoRGB TIFs from C1 - the camera profiles will provide the right input so that there is no risk that the imaginary colours of ProPhoto will be allocated.

It's always nice to learn about the "well behaviour" of certain colour spaces ;-)
« Last Edit: July 13, 2009, 04:14:11 PM by tho_mas » Logged
marcmccalmont
Sr. Member
****
Offline Offline

Posts: 1734



« Reply #11 on: July 13, 2009, 09:16:16 PM »
ReplyReply

You might try PF Killer http://www.dream-fusion.com/mkII/pics/pfkiller.zip
Marc
Logged

Marc McCalmont
Peter_DL
Sr. Member
****
Offline Offline

Posts: 423


« Reply #12 on: July 14, 2009, 05:31:51 PM »
ReplyReply

Quote from: tho_mas
The ProPhotoRGB 72/7/230 is displayed as Lab 20⎢84⎢-125 in the color tool of Photoshop. Nice so far. Now convert your ProPhotoRGB 72/7/230 from ProPhotoRGB to Lab. One would expect that the Lab values remain the same. But they change to 38/55/-90... which is a violett.
ProPhoto: [attachment=15383:prophoto.jpg]
ProPhoto converted to Lab: [attachment=15384:lab.jpg]
The Color Picker in Lab mode (as far as RGB numbers are concerned) refers to the RGB working space as set forth in the Color Settings. If it is different from your source space ProPhoto RGB, same RGB numbers in the Color Picker lead to different Lab readings and vice versa. In other words, so far no color shifts from conversion of ProPhotoRGB 72/7/230 to Lab. Just when you convert to a smaller working space as set forth in the Color Settings  .

Anyway the "blue turns purple" problem is nicely illustrated here.

Peter

--
Logged
tho_mas
Sr. Member
****
Offline Offline

Posts: 1697


« Reply #13 on: July 15, 2009, 04:48:32 AM »
ReplyReply

Quote from: DPL
The Color Picker in Lab mode (as far as RGB numbers are concerned) refers to the RGB working space as set forth in the Color Settings. If it is different from your source space ProPhoto RGB, same RGB numbers in the Color Picker lead to different Lab readings and vice versa. In other words, so far no color shifts from conversion of ProPhotoRGB 72/7/230 to Lab. Just when you convert to a smaller working space as set forth in the Color Settings
Thanks for pointing that out, Peter!
But honestly I'm a bit confused now.
If I create a new document in ProPhoto and afterwards convert to Lab (not via "mode" but in the menu dialogue "convert to profile")... why does the conversion refers to the default working space in the colour settings here? Is this a peculiarity of Lab conversion? Because... converting RGB profiles relative colormetric (either way which ones) the Lab values always remain the same while the RGB values change (sole exception: if the target color space is smaller and therefore clipped colours result in different Lab values). Assigning a profile is the other way around: RGB values remain but Lab values change.
Why do the Lab values change when I convert a RGB file to Lab? I suspect that you are completely right as the violett here shows the desaturation caused by the translation to my working space (ECI-RGB)... but still I don't get it. Thanks!
« Last Edit: July 15, 2009, 04:51:30 AM by tho_mas » Logged
Peter_DL
Sr. Member
****
Offline Offline

Posts: 423


« Reply #14 on: July 15, 2009, 10:29:40 AM »
ReplyReply

Quote from: tho_mas
Why do the Lab values change when I convert a RGB file to Lab?
After conversion (from pRGB) to Lab there is no knowledge anymore about the source space. While the Lab numbers de facto remain unchanged, the RGB readings as given in the Info palette now refer to your standard working space as defined in the Color settings. Note that the conversion from pRGB 72, 7, 230 to Lab 20, 84, -125 is completely reversible. See Info palette.

It’s just the Color Picker which follows a different logic. The RGB combo of your last input (in pRGB) is obviously preserved, but now (after conversion to Lab) refers to your standard working space. Different Lab numbers are indicated accordingly - unless you click on the image to synchronize the Color Picker.

In other words, the Lab values of your image color do not change. It’s just the color ‘in’ the Color Picker which has become different.

Peter

--
« Last Edit: July 15, 2009, 10:36:37 AM by DPL » Logged
tho_mas
Sr. Member
****
Offline Offline

Posts: 1697


« Reply #15 on: July 15, 2009, 11:02:43 AM »
ReplyReply

Quote from: DPL
After conversion (from pRGB) to Lab there is no knowledge anymore about the source space. ...
makes sense - thank you!
Logged
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad