Perceptual Conversion in Working Spaces

[Robert Ardill]

Yes, Robert, you've clarified your point.

Hi Tim,

You don't say if it what I said makes sense and if it does, do you agree?

I must say that I'm getting quite fond of Andrew's Balls.  It's an image that can be quite useful to show how not to do things.

For example, here is a comparison between trying to bring the image into the sRGB gamut by desaturating each of the colors in turn using the Hue/Saturation adjustment in Photoshop (left image) compared with letting the CMM do it using the Perceptual mapping profile (image on the right).



I think it demonstrates quite well how the CMM can do a far better job of converting the colors than we can do manually.  

In order to bring the colors into gamut using the Hue/Saturation adjustment I've ended up with a terribly desaturated image.

What is also interesting is to see what can be done with these images to try to make them better, post conversion to sRGB.  Well, I tried various things on the left image (Hue/Sat and Curves) and I couldn't make any improvement.  On the other hand with the right image, just by applying a small curves adjustment I could improve the image quite a bit.

Here are the two images again, but this time with a small curves adjustment to the right image.  You should be able to see that the image looks more vibrant than the corresponding image above.




Robert

[Eyeball]

I must say that I'm getting quite fond of Andrew's Balls.

Thanks for giving me my first good laugh of the morning, Robert.

I agree with your conclusion regarding global saturation adjustments vs. letting perceptual do its thing.
I think if you're going to do things manually and want better results than the CMS you need to do more localized/non-linear adjustments like using saturation masks or brushing in adjustments for just the problematic areas.

I have had some success in creating my own "pseudo-perceptual" adjustments in Lightroom using custom camera profiles.  I start with my normal custom camera profiles and then tweak the high-saturation points in the DNG Profile Editor.
They seem to work pretty well at taming out-of-gamut colors without impacting in-gamut colors too much.
I haven't done detailed analysis of the results and it is likely lacking for pro work but I mention it in case you haven't tried it and might be interested in experimenting.
If Eric's around, I am sure he could comment regarding possible gotchas for that approach.

[Robert Ardill]

Thanks for giving me my first good laugh of the morning, Robert.

I agree with your conclusion regarding global saturation adjustments vs. letting perceptual do its thing.
I think if you're going to do things manually and want better results than the CMS you need to do more localized/non-linear adjustments like using saturation masks or brushing in adjustments for just the problematic areas.

I have had some success in creating my own "pseudo-perceptual" adjustments in Lightroom using custom camera profiles.  I start with my normal custom camera profiles and then tweak the high-saturation points in the DNG Profile Editor.
They seem to work pretty well at taming out-of-gamut colors without impacting in-gamut colors too much.
I haven't done detailed analysis of the results and it is likely lacking for pro work but I mention it in case you haven't tried it and might be interested in experimenting.
If Eric's around, I am sure he could comment regarding possible gotchas for that approach.

Well, there you go ... at least he's got balls, it's a start, so there has to be hope there.  Generally I've found that it's better not to despair.

We've had this discussion before of course (and no doubt it's all been said many times before). Here for example: http://www.luminous-landscape.com/forum/index.php?topic=90987.msg741400#msg741400

But what you're saying has piqued my interest ... and I have no idea what you're doing.  So I would be VERY interested to find out more, if you have the time, that is.

Robert

[bjanes]

Hi Tim,

You don't say if it what I said makes sense and if it does, do you agree?

I must say that I'm getting quite fond of Andrew's Balls.  It's an image that can be quite useful to show how not to do things.

For example, here is a comparison between trying to bring the image into the sRGB gamut by desaturating each of the colors in turn using the Hue/Saturation adjustment in Photoshop (left image) compared with letting the CMM do it using the Perceptual mapping profile (image on the right).

Robert,

You have made your point very well, but what CMM was used in your example? Smart or Dumb?

Bill

[Eyeball]

It's pretty simple.  I just bring one of my custom camera profiles into the Adobe DNG Profile Editor, select the most saturated points, and reduce the saturation for those points.
There's no reason you couldn't tweak the hue and lightness also if that helped you get things in gamut.
I leave the points closer to the center alone since that is what gives me the non-linear adjustment.  If I tweaked all the points, it would be like your global saturation reduction.

Obviously, just like a perceptual rendering intent I am sacrificing color accuracy for smoother transitions on out-of-gamut gradients.

I have found it works pretty well as an alternative for when I want to convert highly-saturated images to sRGB for web use although I don't claim it is perfect.

Here is a screen-shot from the profile editor that hopefully gives you an idea of what I'm talking about.  For this particular profile I was wanting to tame mainly reds and oranges.



A somewhat similar "pseudo-perceptual" techinique that can be used in PS is to bend down the color curves in LAB.  Here is an example.



Again, I don't claim that it is the best technique around.  I know that some folks don't like to round-trip to LAB and I also know there are probably dozens of alternative techniques for gamut adjustments in PS but I do think it has the advantage of giving you pretty useful visual feedback of where the problematic areas are, how far out they are, and how you might approach a "perceptual" curve to address them.

[Robert Ardill]

Robert,

You have made your point very well, but what CMM was used in your example? Smart or Dumb?

Bill

Hi Bill,

I used the Dumb one - just converted to the Argyll sRGB profile using Perceptual.

There should be relatively little difference between the Smart and Dumb with this image as it is so saturated in ProPhoto that the Smart conversion isn't going to be able to do a lot better than the Dumb one.  Still, here is a comparison, with the Smart at the left and the Dumb at the right:



[Right-click on image to see full size]

At first sight there seems to be little between them, but if you do a difference you can see that there is quite a lot in fact.  Here I've taken two spot readings at the exact same positions (I used a grid in Photoshop to make sure I hit the same exact spots) and as you can see, the two left spots are significantly more saturated that the two right spots: which is what should happen.

Always nice to confirm theory with practice

With less extremely saturated images the improvement of the Smart over the Dumb should be greater.

Robert

[Robert Ardill]

It's pretty simple.  I just bring one of my custom camera profiles into the Adobe DNG Profile Editor, select the most saturated points, and reduce the saturation for those points.
There's no reason you couldn't tweak the hue and lightness also if that helped you get things in gamut.
I leave the points closer to the center alone since that is what gives me the non-linear adjustment.  If I tweaked all the points, it would be like your global saturation reduction.

Obviously, just like a perceptual rendering intent I am sacrificing color accuracy for smoother transitions on out-of-gamut gradients.

I have found it works pretty well as an alternative for when I want to convert highly-saturated images to sRGB for web use although I don't claim it is perfect.

Here is a screen-shot from the profile editor that hopefully gives you an idea of what I'm talking about.  For this particular profile I was wanting to tame mainly reds and oranges.



A somewhat similar "pseudo-perceptual" techinique that can be used in PS is to bend down the color curves in LAB.  Here is an example.



Again, I don't claim that it is the best technique around.  I know that some folks don't like to round-trip to LAB and I also know there are probably dozens of alternative techniques for gamut adjustments in PS but I do think it has the advantage of giving you pretty useful visual feedback of where the problematic areas are, how far out they are, and how you might approach a "perceptual" curve to address them.

It's really interesting to see the different techniques that are being used by different people.  Both of your methods seem worth having a good look at.  The off-the-cuff issue I would have with them is that I don't quite see how you can gauge the adjustments ... except by experience and visually, I guess?  With the Lab version you could soft-proof to sRGB with gamut warning.  But I wasn't too successful with Andrew's Balls ... it would be interesting to see what you can achieve with your experience.

Robert

[Wayne Fox]

I must say that I'm getting quite fond of Andrew's Balls.  It's an image that can be quite useful to show how not to do things.

Actually balls are from Bill Atkinson ..

https://www.dropbox.com/sh/ka35g3gtyd10823/iRGJexS5hA/Twenty-Eight%20Balls.tif.zip

[Robert Ardill]

Here is a suggested workflow to allow the selection of either a Relative or Perceptual intent when exporting photos from Lightroom to sRGB.

• In the develop module soft-proof to decide whether Perceptual is better than Relative.  To do this use the ppRGB-sRGB profile I gave a link to above (or your own if you prefer, of course)
• If you prefer Relative then export the photo to your export folder using sRGB.  If you prefer Perceptual, export using ppRGB-sRGB.  Lightroom always uses a Perceptual intent for exports if the profile is table-based (something Adobe needs to fix IMO).  At any rate this means that the exported image will have been rendered using a Perceptual intent and the embedded profile will be ppRGB-sRGB
• In Photoshop, create an action called AssignSRGB which does the following: Assign sRGB, Save, Close
• Make sure Color Settings is set to [RGB: Preserve Embedded Profiles]
• Run Automate Batch on your export folder with the Action set to AssignSRGB, Suppress Color Warnings checked, Destination None

All the files in your export folder will now have sRGB embedded.

If anyone can think of an easier workflow I would be grateful if you would let me know what it is.

Robert

[hugowolf]

When posting two images for comparison visually, you might want to try mirroring one of them, it saves some of the metamerismic problems associated with the LR siding with LR.

Brian A

[Robert Ardill]

When posting two images for comparison visually, you might want to try mirroring one of them, it saves some of the metamerismic problems associated with the LR siding with LR.

Brian A

Good thinking!