Aspen Yellow and Monitor Gamut

[Downrigger]

This is a view to the northeast, on a partly cloudy morning, of the Capitol Reef area from nearby Boulder Mountain. This and a few other images with aspens at peak on this trip suggest to me that the color gamut on my (calibrated, Mac, some version of slightly bigger than sRGB) monitor is weak in the brilliant lemon yellows that would accord true fidelity to the hues of fall aspens. Comments on that, and general C&C much appreciated.

[Polar01]

Mark I think it is just your WB, I took it into ACR and reset the WB on that foremost Aspen's white trunk, made a big difference on my monitor.

Cheers: Allan

[Downrigger]

[QUOTE=Polar01;474443]Mark I think it is just your WB,

Cheers: Allan[/QUOTE
Thanks, Alan... that helps partly, maybe, but we're sort of going in circles as I had shifted the yellow hues a little away from orange (which was the original problem) towards green, and white balancing off the trunks you have done warms it back slightly towards the original set of hues.

What I am missing here is a lemony bright yellow characteristic of this foliage that is distinct from the orangey/muddy yellows in this image. Pulling the yellows away from orange helped a bit (maybe overdone here) but nothing I do really gets me to where I wish to be. I've done many versions starting with the original RAW and to me, the color I want is simply not there.

It's hard to explain without displaying the yellow I want, which, circularly, I believe to be not available to us on our monitors. Or maybe I'm nuts.

[Downrigger]

Here's a little more on this issue in a timely concurrent thread:

See XpatUSA’s post quoting Eric Chan of Adobe regarding loss of some light yellows when working in sRGB in the thread “Clarification on Color Management (Working Spaces and Rendering Intents).

[ajohnw]

I think your problem is largely down to this



The blue channel. Not just that it's clipping but that cameras seem to have some problems accurately reproducing yellow tones. Red is clipped too. Net result is a colour imbalance that's tough to correct.

Actually there is such a mass of single channel clipping in the shot posted it's hard to see if it's in the leaves or around them. For instance an enlarged view of what is mainly blue clipping. The whole area of leaves looks like this. The clipping is at the dark end - in otherwords higher levels are needed their to register the colours the camera recorded which of course can be incorrect anyway.



John
-

[Downrigger]

Thanks John... very much. But I don't get it. The blue channel is roughly as you show it (extended bimodal) in my untouched RAW, in my finished TIFF and in my exported JPEG, but in none of those is there any clipping at all. I'm not sure why we are seeing such different data, could be user or software error at my end but eludes me what might be...

Even when I push my blue channel darker to create clipping, I get none in the leaves where you are showing it.

[xpatUSA]

Mark, I also find that blue clipping to zero is indicative of out-of-gamut yellows. I have one flower shot that is OK in ProPhoto but when shown in sRGB it has red clipped to 255 and blue clipped to 0. One solution is to reduce overall saturation a bit at a time until both the red and blue channel come back "inside" the color space. There are other ways but they can be painful to use, e.g. ICC V4 sRGB.

I see that your embedded profile is straight sRGB V2. As you know, it is a display class profile and uses a matrix conversion. That means, for viewing on a monitor, you get Relative Colorimetric rendering whether you like it or not, even if it says "perpetual" in your editor.

I've just downloaded a display class profile that promises better because it has color look-up tables in it just like printer ones do. It's called ppRGB-sRGB.icc and I can use it in RawTherapee quite successfully. It might even fix your Aspens to an extent . . .

Interested?

[IzzieK]

I'd like to see the result of what Mark can do with his photo here. Sounds too good to be true. I have a book of Martin Evening called Photoshop for Photographers and he seemed to prefer the ProPhoto RGB more for preserving the real colour of a photo you wanted printed as it has a wider gamut of colours...

[ajohnw]

Thanks John... very much. But I don't get it. The blue channel is roughly as you show it (extended bimodal) in my untouched RAW, in my finished TIFF and in my exported JPEG, but in none of those is there any clipping at all. I'm not sure why we are seeing such different data, could be user or software error at my end but eludes me what might be...

Even when I push my blue channel darker to create clipping, I get none in the leaves where you are showing it.

One of the criticisms of Rawtherapee is that by default it shows single channel clipping as it encourages people to get rid of the lot often leaving a rather flat image. I know for a fact that LR doesn't do this from an image I reprocessed - long yellowish grass. Same comment from the person who posted it - no clipping but RT showed lots.

The clipping levels in RT might not be set at zero which may sound strange but back lighting on panels messes up values which are too low to register colour. Actually I am sure that they are lower than that. Checking they are set to 0 and 255.

I still think the best answer to this problem is to post a raw file on filebin or where ever and hope that some one who can do it mentions what they did. Yellows are a definite problem on digital cameras, I have found an answer in part to bright ones - under expose.

Blue channel problems seem to crop up for me but the main culprits should be red and green as this shows.



But just what mix is golden yellow etc ? I suspect it's all down to the RGGB bayer layout.

John
-

[Simon Garrett]

This is odd! I downloaded the jpeg, and when I open it in Photoshop I can't see any of the clipped pixels ajohnw shows. I mean: they're not clipped for me in Photoshop in any channel. I've checked (as far as I can) a number of the specific pixels that ajohnw shows as white in his post, and for me, all three channels are well below 255.

Also, although there is clipping in the blue channel, it's not in the leaves but in the sky.

It's hard to know what the leaves might have looked like in a wider colour space, as the image posted is in sRGB colour space.

If the original image was in raw, perhaps the OP could post that, or at least post an Adobe RGB jpeg.

[xpatUSA]

I had a look at the image in the OP using ColorThink - a program which can show the color gamut of an image in glorious 3D. It it can also show a color space's gamut. Mark posted his image in sRGB - the right thing to do for the web, of course. Anyhow, here's the image gamut first (it will look funny to some because it's using the Lab model, not CIE xyY):



As suspected, during conversion from camera or from ProPhoto space, a lot of yellows ended color-clipped, see the straight-ish yellow boundary line on the right. By "color-clipped" I do not mean luminance clipped due to over-exposure.

I added in the sRGB color gamut for comparison:




Bingo. Now we can see that many yellows are hard up against the sRGB boundary and this is because those yellows were out-of-gamut so far as sRGB is concerned.

[Downrigger]

Now we can see that many yellows are hard up against the sRGB boundary and this is because those yellows were out-of-gamut so far as sRGB is concerned.

Thanks so much Ted, I'm hard up against some social obligations as the day wears on but I will study this. It looks like what I thought - some aspens at peak are out of the sRGB color space. My memory of that day tells me this and I think that's what you've shown more objectively.

[ajohnw]

Well here's one with a lot more yellow saturation. As much as is possible in sRGB I suspect.



John
-

[Downrigger]

If the original image was in raw, perhaps the OP could post that, or at least post an Adobe RGB jpeg.

Thanks for checking, Simon. I think the Adobe RGB would look the same as what's already up, though, (for most of us) as we are limited to sRGB on our monitors.

[Downrigger]

Thanks, all for the input. I feel as though I have picked up little theoretical and practical insight and deepened my realization that one simply cannot take color for granted. Color on the screen or paper is not often quite the same HSL as on the object as captured... consciously or unconsciously we alter the palette, sometimes not inconsiderably.

John. Thanks. I can get something like that with a further push on saturation and luminance, but the yellow leaves are still dull-seeming to me compared to the lemon glow one sees on the retina and I start to get some edge effects from pushing the pixels around so much. I think Ted's color think analysis is pretty explanatory - I'm simply looking for something that is a bit out of sRGB gamut.

[Downrigger]

It's called ppRGB-sRGB.icc and I can use it in RawTherapee quite successfully. It might even fix your Aspens to an extent . . . Interested?

Thanks, Ted. Out of my personal skill/knowledge gamut - I'd need a little schooling to know how to take advantage of your nice suggestion.

[ajohnw]

I had another play with this image because I don't think it has anything to do with sRGB. One thing that stuck me is that there should be some green in the grass. There is also some in the aspens. So I had a play with rgb levels. Unfortunately the additive colour mixing part of my brain isn't very functional especially with subtle variations. This is the result.



I used the GIMP. Auto has tried to introduce some green. Often if channels aren't clipped this gives a close result. This guy talks slowly and takes his time but gets it across, 3 vids on this sort of problem. Probably applicable to all packages.

https://www.youtube.com/watch?v=zTIBx_JZ3po

On the manual one I tried to get some more realistic greens in the grass where it looks like there should be some. Late so gave up. Same sort of thing as the video plus gamma adjustment on colour channels.

If you look at the aspen yellows they have fairly subtle variations but I feel that the greens are still incorrect. The violetish cast in the sky is reduced on both.

Some one mentioned aRGB. All that will do is give a jpg that is closer to posterisation and will get changed to sRGB on many monitors anyway. i suspect that will alter appearances on a shot like this if in fact it does show any extra out of sRGB gamut colours. There is no way it can do anything else so I wonder why this subject often crops up and is largely ignored by some.

The whole thing has nothing to do with gamut just a weird colour balance problem. One thing I will mention on blue channels in particular is that I have had extreme problems with Nikon shots from raw when a poor camera profile was used. There happened to be a lot of blue in the first shot I processed. Similar things could happen to other channels or mixes intended to produce specific colours. Nikon's raw software seems to make a much better job of this sort of thing immediately.

John
-

[Downrigger]

I don't think it has anything to do with sRGB.
One thing I will mention on blue channels in particular is that I have had extreme problems with Nikon shots from raw when a poor camera profile was used. John
-

Great thought, John. I like your repairs - the second is truest to my recollection, for what that's worth (sometimes I have something other than fidelity in mind, but with the yellow aspens I was hoping for it). As for the grass, it was straw colored, but for that I'm ok with green.

I'd suggest that there is still a little issue with gamut here, and I think Ted's analysis leads us to that conclusion.

I'm not sure what you meant about Nikon RAW files when "a poor camera profile was used", as I understand a RAW file to be indifferent to whatever profile the camera happens to be is set to. If I am incorrect, I need to make a major change in how I go about my captures.

[xpatUSA]

Hello Mark,

I had a go from an analytical point of view, using the following steps:

I looked up "lemon yellow" and chose HSV = 51deg, 82%, 95% from here: http://www.perbang.dk/rgb/F3D52C/ although it perhaps should be a little more toward green?

I opened your original in RawTherapee and scanned the picker around in that brighter, smaller tree at center. Reds ranged from 96-99, greens from 75-82, blues from 0-60 but generally more toward zero, so the yellows tend toward saturation. The hue was about 49 deg. perhaps a little too far toward red?

I adjusted the color curves as follows - without much regard for anything but the aspens.

Added more blue in the shadows to lighten the leaves and reduce (yes) the saturation (the less saturation, the lighter the color):



Now the hue is still toward red, so added some green in the highs:



Now the hue had gone too far toward green (perhaps should have left it) so added a bit more red to both shift the hue and up the brightness (value, not lightness):



After all that, the trees looked slightly dull so I upped the chrominance in Lab with 'avoid hue shift selected':



No adjustments were drastic and there was headroom for a bit more hue shift away from red, and slightly more saturation.

In FastStone Viewer comparison is easy:



Differences are subtle. A blue cast on the trunks is noticeable - but color-balancing them to neutral would undo all of the above. Slight HSV equalization in RawTherapee might help a bit.

In the original at left, when comparing side by side, there appears to be a very slight hue shift towards red in the sky whereas at right the sky is more toward cyan.

[ajohnw]

A raw converter uses a camera profile just as it's internal jpg engine in the camera does. It scales r g b values and produces something suitable for viewing rather than just providing the linear data from the sensor. It's a little like gamma being added to images that are to be displayed.

If your using Adobe stuff there will be adobe standard and other profiles such as portrait and landscape tucked away some where on the machine and the ability to change from one to another in the package. They all provide different r g b scalings from sensor to view. Some packages have the ability to not auto brighten as well but I only know of one.

As an alternative some packages provide something usually called colormatrix for when a camera profile isn't available. Names might vary. It's a generic sort of thing and seems to be an absolute disaster on some cameras. Packages usually provide correction sliders to alter the r,g,b balance but they are complicated to use.

I use rawtherapee a lot now for raw. On that I have to go into one of the colour panels and select the correct profile or it uses colourmatrix. There seems to be an auto find option but I haven't tried to get it to work as yet. Profiles may need to be in a specific directory. I'd guess Adobe stuff looks after this by loading Adobe Standard for the camera automatically - still leaving the others to be selected if needed or maybe it reads the camera setting from the exif.



On the greens I meant some parts of the grass not all. Also the green in some of the aspens. We do have aspens over here.



Good luck with bigger gamuts but I feel it's a red herring. Mainly of use for prints where the dynamic range is rather limited. I know of one aRGB colour that causes sRGB a lot of grief - a shade of teal. I ran aRGB test colours through my colorimeter.

John
-