Ad
Ad
Ad
Pages: [1] 2 3 »   Bottom of Page
Print
Author Topic: Raw converter that supports CIELab?  (Read 5316 times)
MirekElsner
Jr. Member
**
Offline Offline

Posts: 75


WWW
« on: December 31, 2013, 09:14:26 PM »
ReplyReply

As weird as it sounds, I am looking for a raw converter that supports direct conversion to Lab or can capture colors that are outside of ProPhoto RGB gamut. Is there any?
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #1 on: January 01, 2014, 06:45:41 AM »
ReplyReply

As weird as it sounds, I am looking for a raw converter that supports direct conversion to Lab or can capture colors that are outside of ProPhoto RGB gamut. Is there any?

The more recent versions of Adobe Camera Raw can render into L*a*b but do so from their internal working space which uses ProphotoRGB primaries and the resulting output can not exceed the Prophoto gamut. According to Bruce Lindbloom, ProphotoRGB covers 91.2% of the L*a*b gamut. This is more than sufficient for practical work, since no currently available output device can come anywhere near to reproducing the entire L*a*b gamut. Furthermore, ProphotoRGB exceeds the gamut of real world surface colors (reflective colors seen in nature, see page 10 of this document). Additional gamut would be needed to encode emissive sources such as neon lights.

Interestingly, integer encoding of L*a*b as done in Photoshop and many other applications can encode only 97% of the L*a*b gamut, and I think that it is unlikely that any common raw converter could represent the entire L*a*b gamut.

Regards,

Bill



Logged
digitaldog
Sr. Member
****
Online Online

Posts: 8610



WWW
« Reply #2 on: January 01, 2014, 12:15:58 PM »
ReplyReply

I suspect every raw converter at some point starts with an RGB assumption going out to whatever. So a direct conversion to Lab from what?
Logged

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

Posts: 1264


« Reply #3 on: January 01, 2014, 01:02:12 PM »
ReplyReply

the gamut of real world surface colors (reflective colors seen in nature
but real world != reflective colors seen in nature... because nowadays you have to deal w/ artificial ones (even not emissive).
Logged
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1264


« Reply #4 on: January 01, 2014, 01:02:52 PM »
ReplyReply

As weird as it sounds, I am looking for a raw converter that supports direct conversion to Lab or can capture colors that are outside of ProPhoto RGB gamut. Is there any?
try rpp
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #5 on: January 01, 2014, 01:26:51 PM »
ReplyReply

I suspect every raw converter at some point starts with an RGB assumption going out to whatever. So a direct conversion to Lab from what?

Since the Bayer CFA is an RGB device, that sounds like a reasonable assumption. Chapter 6 of the Adobe DNG specification deals with mapping of the camera color space to CIE XYZ, suggesting that this mapping is performed in Adobe raw conversion. As I understand things, CIE Lab is a mathematical derivation of CIE XYZ, and the Lab values could be derived with no loss of data from the XYZ values.

Bill
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3443


« Reply #6 on: January 01, 2014, 01:35:28 PM »
ReplyReply

I suspect every raw converter at some point starts with an RGB assumption going out to whatever. So a direct conversion to Lab from what?

Exactly. It might help if the OP explained why he thinks that direct CIELab encoding offers a benefit, maybe he's after something that's useful, but I have my doubts. Even conversion from camera RGB space to CIELAB would require floating point numbers to avoid quantization steps that are too large to be accurate.

Cheers,
Bart
Logged
MirekElsner
Jr. Member
**
Offline Offline

Posts: 75


WWW
« Reply #7 on: January 02, 2014, 01:36:11 AM »
ReplyReply

Exactly. It might help if the OP explained why he thinks that direct CIELab encoding offers a benefit, maybe he's after something that's useful, but I have my doubts. Even conversion from camera RGB space to CIELAB would require floating point numbers to avoid quantization steps that are too large to be accurate.

Cheers,
Bart

There is no practical reason. I just want to see what colors is my camera (after post processing) able to produce. I know there are at least some colors outside of ProPhoto, because I can see some clipping when looking at some color plots.

The attached image shows a lot of colors from a shot of some trees (the picture is mostly green) into xyY coordinates. This is a ProPhotoRGB image (from LR) and the red triangle shows boundaries of the embedded profile. There is some clipping of the greens, less extensive but similar as what I see with plots made from images converted from ProPhoto to colorspaces like AdobeRGB, so I guess there is more to see if I use appropriate converter (and more colorful captures).
« Last Edit: January 02, 2014, 01:44:09 AM by MirekElsner » Logged
MirekElsner
Jr. Member
**
Offline Offline

Posts: 75


WWW
« Reply #8 on: January 02, 2014, 01:58:05 AM »
ReplyReply

try rpp

Thanks!
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3443


« Reply #9 on: January 02, 2014, 03:14:19 AM »
ReplyReply

There is no practical reason. I just want to see what colors is my camera (after post processing) able to produce. I know there are at least some colors outside of ProPhoto, because I can see some clipping when looking at some color plots.

Hi Mirek,

I understand. Do realize though, that the color plots are based on approximated RGB coordinates that change when e.g. a lower saturation is used during Raw conversion. You are almost certainly not looking at spectrophotometrically correct color coordinate positions. Also, the larger the gamut is, the larger the quantization steps between subsequent integer number coordinates (that's why I mentioned that floating point numbers would be required). Also, most actual images do not fully use the full gamut space, and in fact Prophoto RGB also has the room to encode non-existing colors that will therefore never be used.

Quote
The attached image shows a lot of colors from a shot of some trees (the picture is mostly green) into xyY coordinates. This is a ProPhotoRGB image (from LR) and the red triangle shows boundaries of the embedded profile. There is some clipping of the greens, less extensive but similar as what I see with plots made from images converted from ProPhoto to colorspaces like AdobeRGB, so I guess there is more to see if I use appropriate converter (and more colorful captures).

Yes, good example, although the xyY plot only shows the coordinates at a certain Luminance level, and the space is not perceptually uniform (the issue may more, or less, significant when shown in e.g the CIE 1976 (L*, u*, v*) color space), especially in 3D.

What you probably want, and you'd not be the only one, is a better Saturation clipping indicator (not only that it's clipping, but also by how much), and tools to convert between (or specifically towards output) colorspaces.

As other threads here on LuLa and elsewhere have demonstrated, the Raw data may be unclipped before White-balancing, but combined with a given exposure level that data may become Out-Of-Gamut (OOG) after white-balancing. So the actual Raw conversion may already address some OOG issues later in the workflow, without the need for a wider gamut colorspace.

The drawbacks of an 'Mega' large colorspace (e.g. posterization risk with large manipulations and inaccurate colors due to sparse linear quantization with integer coordinates), may outweigh the benefits for a few colors that could be addressed/tamed in a different way.

Cheers,
Bart
Logged
MirekElsner
Jr. Member
**
Offline Offline

Posts: 75


WWW
« Reply #10 on: January 02, 2014, 02:07:25 PM »
ReplyReply

Do realize though, that the color plots are based on approximated RGB coordinates that change when e.g. a lower saturation is used during Raw conversion. You are almost certainly not looking at spectrophotometrically correct color coordinate positions. Also, the larger the gamut is, the larger the quantization steps between subsequent integer number coordinates (that's why I mentioned that floating point numbers would be required). Also, most actual images do not fully use the full gamut space, and in fact Prophoto RGB also has the room to encode non-existing colors that will therefore never be used.

That's fine, I am interested in data after post processing into visually compelling image. That may theoretically involve not only WB/exposure and other changes in the raw converter, but also other accuracy crippling factors like use of polarizer, redhancer or GND filters.

Quote
Yes, good example, although the xyY plot only shows the coordinates at a certain Luminance level, and the space is not perceptually uniform (the issue may more, or less, significant when shown in e.g the CIE 1976 (L*, u*, v*) color space), especially in 3D.

This particular plot shows all Y levels. I convert the RGB to XYZ and then to xyY and plot all unique colors I found in the image. So a single point in the diagram quite often has multiple stacked values with higher Y on the top. The points are plotted with their original color value, so some hint of Y can be seen from their actual luminosity. Basically, you are looking at xyY 3d diagram from the top.

Quote
What you probably want, and you'd not be the only one, is a better Saturation clipping indicator (not only that it's clipping, but also by how much), and tools to convert between (or specifically towards output) colorspaces.

Yes, that is what I am trying to accomplish. I started with visualization of the data in 2d, next step is 3d. I am thinking about visualizing ∆E as well.  And while I am working on it, I noticed this clipping in ProPhotoRGB.  My tool is RGB-based and any image non-RGB image opened in it must be converted to RGB. That's why I am trying understand what I am potentially throwing away if I open an image in something bigger, that means, possibly a Lab image.

I wanted to point out two things in the attached screenshot:

- The clipping so far always occurs on the right side of the diagram, where it coincides with boundary of the spectral locus. Output gamuts of 2014 aside, I am not sure how practical it would be to salvage these.
- There are two areas where there are visible colors outside of ProPhoto RGB - greens and purples. The areas are not small. I wonder what colors would fit there, I could not find any, yet.

Thanks,
m
Logged
digitaldog
Sr. Member
****
Online Online

Posts: 8610



WWW
« Reply #11 on: January 02, 2014, 02:11:43 PM »
ReplyReply

Can you show a screen capture of the histogram of that image within ACR or LR.
Logged

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

Posts: 8610



WWW
« Reply #12 on: January 02, 2014, 02:31:40 PM »
ReplyReply

The more recent versions of Adobe Camera Raw can render into L*a*b...
Shame we can't do that within Lightroom.

I can produce a histogram that shows no clipping in ProPhoto RGB that then does in Lab within ACR. That seems to illustrate the potential problems with color spaces that encode 'colors' we can't see like ProPhoto RGB. Yank on Saturation in ACR and depending on the image, you can define something we can't see  Roll Eyes
In ACR, open Workflow options and toggle while viewing the histogram update.
Logged

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

Posts: 75


WWW
« Reply #13 on: January 02, 2014, 03:04:43 PM »
ReplyReply

Can you show a screen capture of the histogram of that image within ACR or LR.

Sure, I'll try to attach LR raw histogram, LR gamut warning preview with ProPhoto RGB soft proof and PS histogram of the actual tiff that was used for the plot. It crashes on me so I will try one by one. Just yesterday I browsed through The Digital Print from Jeff Schewe and he had a side note there indicating that according to Eric Chan the LR histogram does not show everything. Just saying...
Logged
MirekElsner
Jr. Member
**
Offline Offline

Posts: 75


WWW
« Reply #14 on: January 02, 2014, 03:06:57 PM »
ReplyReply

Now the soft proof...
Logged
MirekElsner
Jr. Member
**
Offline Offline

Posts: 75


WWW
« Reply #15 on: January 02, 2014, 03:08:47 PM »
ReplyReply

And PS histogram...
Logged
digitaldog
Sr. Member
****
Online Online

Posts: 8610



WWW
« Reply #16 on: January 02, 2014, 03:09:27 PM »
ReplyReply

What you want to do is save those settings for the raw, open in ACR if you have it. Should appear the same. Set workflow options for Lab. Do you see clipping?
Logged

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

Posts: 2756



« Reply #17 on: January 02, 2014, 03:15:32 PM »
ReplyReply

Shame we can't do that within Lightroom.

I can produce a histogram that shows no clipping in ProPhoto RGB that then does in Lab within ACR. That seems to illustrate the potential problems with color spaces that encode 'colors' we can't see like ProPhoto RGB. Yank on Saturation in ACR and depending on the image, you can define something we can't see  Roll Eyes
In ACR, open Workflow options and toggle while viewing the histogram update.

Here is one such example where there is clipping when rendering into Lab but none in ProPhotoRGB

Bill
Logged
MirekElsner
Jr. Member
**
Offline Offline

Posts: 75


WWW
« Reply #18 on: January 02, 2014, 04:01:03 PM »
ReplyReply

What you want to do is save those settings for the raw, open in ACR if you have it. Should appear the same. Set workflow options for Lab. Do you see clipping?

Here we go. How does the Lab gamut look like, does it follow the spectral locus?

Btw, I did couple more experiments. I looked at the image in raw digger and the raw has clipping in the B channel. I wonder why it does not appear in LR... I also tried RPP. It would probably take some time to tame it, but at least I managed to see the histogram - it shows clipping in green and after WB also in reds.
« Last Edit: January 02, 2014, 04:28:36 PM by MirekElsner » Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2756



« Reply #19 on: January 02, 2014, 05:42:02 PM »
ReplyReply

And PS histogram...

You need to refresh the histogram of this image as indicated by the ! in the triangle.

Bill
Logged
Pages: [1] 2 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad