Conversion to HDR is not scaling our SDR Point Curves.

[C.Cella]

I often now find myself doing HDR versions of my already edited SDR images but alas our SDR Point Curves are not adapted/scaled properly to HDR when we covert our images to this mode.

Here is a simple example.

This is the SDR curve.

   

This is the curve when we convert to HDR

. 

This is totally wrong, the curve has not been adapted to HDR ! 

The Highlights and Whites region are lost in the HDR editing, still "crushed" to SDR and we must correct the curve ourselves.

This is the correct result.

  

• So when converting to HDR all our SDR Point Curves should be scaled to HDR and not be kept to SDR.

When going from HDR to SDR the existing point curves should be converted as well.

.

[Rikk Flohr: Photography]

Do you see the same result in Camera Raw or only in Lightroom Classic?

[C.Cella]

@Rikk Flohr: Photography Yes the issue is present in ACR.

The origin of the problem is in the way Camera Raw engine does the conversion to HDR

The  ExtendedToneCurvePV2012 and ExtendedMainCurve are never "converted" from SDR to HDR (or "reverted" form HDR to SDR)

The issue is affecting masking as well.

So I have a done in LrC 12 a Sky Mask and have edited it with a curve.

All works in SDR but when activating HDR happens that mask no longer works BECAUSE there is no "conversion" of the ExtendedMainCurve to the HDR area.

In the screenshots below I visualize the HDR Ranges  (curve not converted to HDR)

That Sky mask has zero High Dynamic Range now, it is still confined to the SD Range so the sky itself looks flat, a blotchy white area with zero transitions AND I msut re-edit it.

Instead with the curve correctly converted to HDR I get these ranges.

Sky shows depth now in HDR, is very tri-dimensional.

.

[DdeGannes]

See the screen capture concerning known bugs with LrC 13.0 and fixes LrC 13.0.1.

[C.Cella]

@DdeGannes 

It's none of those bugs.

This is a "problem" of the HDR conversion.

Let me elaborate more.

1. This is the edited SDR image.

 

2. This is what happens when we convert to HDR.

The image is not HDR at all, as the histogram shows.

3. To make a truly HDR image we must correct the curves and make them HDR.

The math is simple BUT it actually takes times to get the HDR curve...it is not a one click think AND if one has curves in masking is even more time consuming.

The result after the conversion of the curve is this.

 Image is too dark, obviously so because that image was edited for SDR and it doens't encompass all the HDRe yet.

As last step the user needs to do increase exposure to spread the hsitogram to HDR region.

• If the engine was actually converting curves to from SDR to HDR all we had to get a true HDR version of our SDR images is simply increase exposure.

Change ONE SLIDER vs edit all curves.

.

Unfortunately the screenshots are all in SDR so the last image looks with blown out higlights/whites but is not the case in HDR.

I hope the examples help.

[MadManChan2000 · Accepted solution]

You have a misunderstanding of how the point curve works in SDR/HDR.  Curves are never "converted" between SDR and HDR.  It's the same curve.  In SDR, the point curve has values from 0 to 255.  In HDR, it's the same curve, but you now have values from 0 to 500.  The new range for HDR is 256 to 500 and corresponds from SDR white to +4 stops above SDR white. This is all intentional and by design.

Curves optimized for SDR (e.g., big rolloff in highlights) will probably look bad for HDR.  And curves optimized for HDR (e.g., extended highlights) will probably look bad for SDR.  This is normal and expected.  It is also why the HDR button is available at the top of the Edit/Develop stack: it's meant to be used first or early in the workflow, before the use of curves or other features that depend on the SDR/HDR state.  If you start with Curves or other controls and then toggle HDR button, you will likely need to revisit those adjustments.

[C.Cella]

@MadManChan2000 

I understand how it works but I have thousands of SDR images that could be turned into perfectly good HDR with little effort IF we could "convert" the Current SDR Control Points into HDR.

For instance I just opened a random image I edited in SDR time ago.

1.  I have this S curve

2. I activated HDR.

The curve in the SDR region is fine but in HDR is not good so I need covert it to HDR...and I do get the exact same shaped curve, perfectly mathematically scaled to HDR region but I must do it manually (I am not  intelligent enough to do it perfectly without calculator so it takes time)

3. At this point all I needed to do was increase exposure by +1.3 Stop.

Result: image is a perfectly good HDR version of my SDR edit, nothing looked horrible or wrong, all matched.

So I do have to revisit my images BUT it is unnecessary long.

Re-editing curves for HDR is very time consuming, with a direct "SDR to HDR curve conversion" time and effort would be spared. 

Here in the link the image of the Cat I used  as example, there are 3 snapshots that show the different approaches.

https://we.tl/t-NZD2qEqjHb

.

[Jao vdL]

I don't think you want to mathematically completely scale the curve to the HDR range like you are doing. That makes your image look crazy. HDR is about extra headroom, not about the main part of the image being brighter. Most of the image should look the same in SDR as in HDR, just the highlights should pop out into the HDR region. The way you are editing, you're making the entire image incredibly bright. This is not what should happen. In the cat image for example, you just want to get more detail in the bright area top left and in the halo around the cat. The cat itself should remain the same brightness. So it seems to me ACR/Lightroom does the correct thing by just linearly extending the curve instead of scaling it to the entire region. The last cat image for example looks wrong even accounting for the image being blown out in the screenshot. You do not want to do it that way. Only the highlights should be blown out, not the main part of the image. What kind of display are you editing on? 

Also , you should realize that the highlight recovery that ACR always does will have a different effect in HDR because it doesn't have to do as much when there is more headroom, so an image that appears to have nothing blown out  when edited in SDR will often have actual data in the HDR range when you switch. This is normal and due to how ACR tries to fit everything in the SDR range even if your highlights and whites sliders are at zero. 

[JohanElzenga]

I completely agree. Read the great blog that @MadManChan2000 wrote about HDR editing (https://blog.adobe.com/en/publish/2023/10/10/hdr-explained), where he says that "sunglasses are not the answer". Cats do not emit light, but your screenshot suggests that in your photo the cat looks like it does...

[C.Cella]

@JohanElzenga download the file.

I already said the screenshots are SDR so it's not the reference to look at.

Also you could have looked at the histogram in my screenshots,  nothing is clipping.

@MadManChan2000 article is valid for HDR editing from the ground up.

But we all have thousands of images that have been edited for SDR and now need to be "converted", re-edited.

In a quest for efficency, in the contest of re-editing scaling curves helps.

The math to achieve the desired curve scaling is beyond trivial and I think a quick "scale curve" option could be easily doable.

[JohanElzenga]

@JohanElzenga download the file.

I already said the screenshots are SDR so it's not the reference to look at.

Also you could have looked at the histogram in my screenshots,  nothing is clipping.

By @C.Cella

Screenshots are always SDR, but that does not mean you can't use them to get an idea of what the HDR image looks like. Anything that is clipped in the SDR screenshot will be brighter than paper white in HDR, meaning it will look like it is emitting light rather than reflecting light. I'll download your image later (working on my iPad right now), but I'm convinced that will not make me change my mind. The comments from @Jao vdL only confirm what I think I am seeing. The fact that your histogram isn't clipped does not mean anything. The fact that half the histogram is in the HDR range does, however.

I have been playing quite a bit with older images and I agree with what Jao says. HDR is not about making images look 'normal' on a super bright screen. HDR is about using the headroom that the image has but cannot be shown on a SDR display.

[C.Cella]

Clearly the message is not coming through.

This is not about how to do HDR.

That cat photo is not going to be shown on a screen at the Moma museum.

It was just an example.

The photo is not even mine.

@MadManChan2000 you have made clear that it was never considered to scale curves when HDR is enabled because: "Curves optimized for SDR (e.g., big rolloff in highlights) will probably look bad for HDR." 

Fine by me.

No oppositions to that.

I just find that scaling curves + changing exposure can actually provide a valid method to go from a SDR image to an HDR one.

So.

Should I feature request that or this thread can be turned into a fully fledged Feature Request !?

.

[C.Cella]

@Jao vdL did you download the file?

Because you will clearly see that I was able to go from SDR to HDR and nothing breaks.

Whether you like or not that look is subjective and I did it only to illustrate.

Spreading all the pixels across the range is much better than only move some of them.

The approach you suggest, only reassign the highlights and whites to HDR area doesn't  produce good results for me.

I already have used this method on random images and by scaling the curves to HDR and changing Exposure I get a perfectly usable conversion.

Only problem is scaling the control points in the curve is not fast.

I am editing on MacBook Pro M1 Max.

.

[Jao vdL]

Ah I edit on a M1 Max too using the built-in display for HDR. What you are doing in your method is simply using the display as a very bright SDR display. You are not using it as a HDR display with an extended range of brightness. You simply made everything louder (this one goes to 11!). The result is exactly as if you just used a much brighter display. That's not the point of HDR.

Of course if you just want to create an extremely loud image that is fine but the HDR converted snapshot looks terrible to me. All it is is an image that was brightened way too much and it looks out of place with every other image on the same display. That is logical because all you did is stretch the entire image to the new maximum brightness instead of use the bigger available brightness range.

I would recommend you watch some of Greg Benz's excellent videos on how HDR actually works here: https://www.youtube.com/watch?v=6V1edlS0O5k

[C.Cella]

@Jao vdL

No offence but you are not getting the point.

The point is to have the ability to scale curves because there will be cases in which that simple scaling does half of the job.

As of now the user can't do it fast and efficiently.

I am for enabling workflows not preventing them because of some rule.

Photography is not a sport, there are no rules.

.

[Jao vdL]

Ah yeah I have no problem with that. As I said, if you want to take it to eleven that is fine. It's going to be a pretty rare thing so should not be default.

That said, if you use the parametric curve instead of point curves to create a similar tone curve, it does do what you want to happen for the point curve so that is a bit inconsistent I would agree.  

[JohanElzenga]

I think the real point is this: nobody is stopping you doing what you want to do. If you want to use HDR for something it was not designed for, then by all means do it. Indeed there are no rules. But do not expect Adobe to spend developer time and resources to create some special 'hdr curve conversion' option that would not make sense when you use HDR editing the way it was designed, just because you would like that for what you like to do. That is not a realistic expectation.

[C.Cella]

---

@JohanElzenga wrote:

I think the real point is this: nobody is stopping you doing what you want to do. If you want to use HDR for something it was not designed for, then by all means do it. Indeed there are no rules. But do not expect Adobe to spend developer time and resources to create some special 'hdr curve conversion' option that would not make sense when you use HDR editing the way it was designed, just because you would like that for what you like to do. That is not a realistic expectation.

---

@JohanElzenga  I am not doing any exotic workflow.

Far from it.

Anyone can take ANY existing SDR curve and "scale it", replicate it expanded to HDR already ...just not in one click, and that's the problem I am having.

 The workflow is perfectly legit.

Suggesting that HDR was not designed for this is, well, not true because we can do it.

The math behind scaling curves is trivial.

I would create a Plug-in to do the "scale curve"  but I do not posses the skills for that.

@johnrellis is the one person I know that could probably implement a "Scale Curve to HDR" option BUT I think the team could as well consider implementing it, alongside the ability to place the extra control point in SDR without having to enable HDR first.

Users have asked for more control points for years.

.

[JohanElzenga]

I think it's pretty useless to continue this discussion, because this is becoming like a scratched record, playing the same tune over and over again. If you want to file this as a feature request, then nobody will stop you or delete the request again. I still think the extended curve was not designed so you can 'scale' an SDR curve to an HDR curve. The fact that you can do this means nothing. Cars are not designed to drive off a cliff, but you can do this. Hammers are not designed to drive a screw into a plank, but hit the screw hard enough and you can do this. Kitchen knives are not designed to kill people, but... I think you are getting my point. File your request and let's start doing something useful again.

[C.Cella]

I submitted the FR.

.