Non-blown highlights bcome blown

[Linwood Ferguson]

Someone mentioned blown hightloghts on a shot I posted. I looked in lightroom and they were not blown.

So I experimented a bit -- I exported to JPG - blown.

I exported to TIF (16 bit) - blown.

I would have expected even as it does color space conversion that it would not shift so as to blow out one channel (red, by the way).

Should I have expected this?

The Lightroom (right) and Photoshop (left) histograms are in the attachments.

Here's the original shot by the way:

[img width=6'' height=399]http://captivephotons.com/Nature/Other-Nature-Areas/Flock-of-Black-Skillers-near/749228'44_xnnEs-L.jpg[/img]

 

[dj_paige]

Are you exporting to JPG using color space sRGB, or some other color space?

 

[Brad Snyder]

Linwood, I did a little refresher reading on this, and I'm not sure I understand what I know. I guess the first question is what color space you exported to? Presuming sRGB, it's clear that sRGB's gamut is a good deal smaller than Lr's Melissa ProPhoto. So you would expect to see some differences. What I can't define is how luminance affects those specific gamuts, so I'm not sure if it's natural for the highlights to be out of gamut.

Two thoughts:
1)Since you have PS, have a look at the soft-proofing, and see if the gamut warning gives you any useful information re: the highlight clipping.
2) I am reliably informed that the Export module of Lr uses Relative Colorimetric rendering intent. You might try the Print to JPG function, where you should be able to select Perceptual intent, which may give you a less obtrusive mismatch in the highlights.

In the past we've also noted a bit a wacky rendering with these specific ranges of sunset oranges, and I honestly am not sure where that stands without more research, which I'm too lazy to pursue right now. I think Victoria had a handle on this, maybe if she stops by, she can remember where it stands. I think that was specifically an internal Lr issue, not export related, but.....

As always, color management issues are a pretty slippery slope to step out onto.

 

[Bosse]

Linnwood,
A JPG-file is just a container of your original file. It can only be in 8bit but in different colorspaces. A TIF-file can be in 8bit or 16bit and in different colorspaces.
What colorspace did you export the JPG-file in? And the TIF-file?

LR uses a widegamut colorspace (almost like ProPhotoRGB) and looking at your LR-histogram there are no blown high-lights. But that is in a widegamut RGB. Eporting to a relatively small gamut RGB like sRGB can produce blown high-lights. Try exporting your photo 3 times into ProPhoto, AdobeRGB resp sRGB. Do you get blown high-lights in all 3?

You wrote: " it would not shift so as to blow out one channel (red, by the way)".
It is very possible that moving from the wide-gamut space to a smaller will blow out one or more channels if that is where the "problemcolors" are. Maybe you have some red colors that are outside of what a small colorspace can show.

Unfortunately there is no way in LR to "softproof" what will happen to your colors once exported. You can do that in AdobeCameraRAW (ACR). Open your RAW-file in ACR. At the bottom you have a blue line (at least in CS4 - I cannot remember CS3) you can klick and choose different colorspaces (among other things). Choose another one then the original, klick OK and observe the histogram. Then do the same with another colorspace and see what happens. In sRGB your photo will probably be clipped, possibly not in AdobeRGB and not if you export as ProPhotoRGB. Then you have to deal with the gamutproblem inside Photoshop.

Brad,

In response to your 2 thoughts:
1) If Linnwood wants to softproof within PS he should export as ProPhotoRGB and then softproff vs AdobeRGB and sRGB. Exporting to sRGB the highlights will probably be clipped allready and then no gamutwarning when comparing to a colorspace of the same gamut or larger.
2) I can be wrong but I have been told that when converting between RGB colorspace, Relative rendering is always used no matter what you choose. Correct me if I am wrong.

Yep, colormanagement is slippery.


// Bosse

 

[Linwood Ferguson]

Have obligations this afternoon and will experiment later, but the short answer is the JPG was exported in sRGB and the 16 bit TIF was exported in Adobe RGB. The Adobe RGB TIF was the one I posted the CS3 histogram from. I would have thought the Adobe RGB was plenty wide not to truncate. I confirmed it is indeed in Adobe RGB by checking in Photoshop. I looked at the TIF in IFranview as well, and its histogram for red shows a bit spike at the right side.

 

[Bosse]

There are colors that are inside ProPhoto gamut but outside of even a large space like AdobeRGB. I found this when I googled for "Prophoto vs AdobeRGB"
http://www.naturephotographers.net/articles12'3/mh12'3-1.html
The bird in the photo has colors not to far away from yours.

I also found a link to this PDF written by gurus Jeff Schewe and Bruce Fraser:
http://www.adobe.com/digitalimag/pdfs/color_managed_raw_workflow.pdf
Some good info in there.

// Bosse

 

[Linwood Ferguson]

I did quite a bit of experimenting.

LR 3 beta - same.

I exported the same image to 16 bit TIF's in sRgb, AdobeRGB and ProPhoto. The histograms from CS3 are attached, in that order left to right.

I naively thought that on export/conversion, if highlights were not blown, they would continue not to be blown. THat somehow it would scale.

I assumed that most of the lack of soft proofing was (Say) a particularly deep purple came out slightly the wrong color. I never expected whole areas of a shot that had detail on my screen to suddenly vanish into a uniform blown color.

But even now that I realize that is not true, the more surprising thing is that AdobeRGB is almost the same as sRgb -- look at the relative shift of the peaks between the three -- sRgb and Adobe are pretty similar (not in vertical scale, but in position -- i.e. how close to the extremes).

I guess the message here is you can't trust the LR Histogram to really tell you if your highlights are blown, if the ultimate target is anything but ProPhoto?

And now I start to understand more why people complain about the lack of soft proofing.

Oh... I also did the conversion in CS3 -- same results. I tried all four rendering intents, same basic results.

 

[Bosse]

Long answer - sorry.
Your comments in italic - my answers below each.

LR 3 beta - same.
I haven´t found anything that suggests that this has changed from 2.6 >3

I exported the same image to 16 bit TIF's in sRgb, AdobeRGB and ProPhoto. The histograms from CS3 are attached, in that order left to right.
First: the exclamationmark inside the triangle shows that the histogram is not updated. Klick on it (or the arrowcircle) to update (but I don´t think it will change so much to make the answers different in this question.
Second: There is a little spike in the Pro-histogram on the far right. Visible in the red channel mostly. That spike can be seen in the LR-histogram as well. There is no way while in LR to tell that that spike is out-of-gamut when converting to a smaller space. Or where the OOG-colors start. Sad but…

I naively thought that on export/conversion, if highlights were not blown, they would continue not to be blown. THat somehow it would scale.
It doesn´t scale - it chops/cuts off. Here is a quote I found when googling: "Relative Colorimetric. When converting between matrix-based profiles, any colors that don't fit in the target profile simply get chopped off. Working space profiles such as Adobe RGB and sRGB are matrix-based as are most input and display profiles."
Read more here: http://www.earthboundlight.com/phototips/matrix-profiles-perceptual.html

I assumed that most of the lack of soft proofing was (Say) a particularly deep purple came out slightly the wrong color. I never expected whole areas of a shot that had detail on my screen to suddenly vanish into a uniform blown color.
It can be both. I have a very good (bad?) printed example which I cannot show here but it is of a flower with bright yellowish-orange-reddish colors. At least they used to be before they got printed on paper. Now they are just a big blob of a mix of those colors. Awful. That can happen anywhere going from a large space to a smaller.

But even now that I realize that is not true, the more surprising thing is that AdobeRGB is almost the same as sRgb -- look at the relative shift of the peaks between the three -- sRgb and Adobe are pretty similar (not in vertical scale, but in position -- i.e. how close to the extremes).
The vertical scale only shows the number of pixels at that position.
I see a position-difference in the red channel. If you compare the peaks in the three red-channels the sRGB is cut off much earlier then the AdobeRGB which in turn is cut off much earlier then the ProRGB.

I guess the message here is you can't trust the LR Histogram to really tell you if your highlights are blown, if the ultimate target is anything but ProPhoto?
True. After a while you learn when you have a "problem"-photo and export that as ProPhoto instead of AdobeRGB or smaller.

And now I start to understand more why people complain about the lack of soft proofing.
Did you try to look at the histogram changing inside AdobeCameraRAW? (that is one of the very few reasons I would go via ACR instead of Developmodule inside LR). If I have a "problem" photo I always export it as a ProPhoto. Then in PS I use softproof (against AdobeRGB since that is my chosen working colorspace) and for example a Hue/saturation adjustmentlayer to bring the faulty nuances into AdobeRGB. Here my very good (and expensive EIZO CG243) can show me what is happening since that monitor can show me almost 1''% AdobeRGB.

Oh... I also did the conversion in CS3 -- same results. I tried all four rendering intents, same basic results.
Choosing different rendering intents when converting between RGB-space makes no difference. You always use relative calorimetric (at least according to the link above)

// Bosse