Tone curve and luminance?

[bdosserman]

Hi all,
   Sorry if this is a dumb question, but I just got LR4, and while the shadows slider seems to work very nicely, I am disappointed with the tone curve. Specifically, it looks to me as though the curve is being applied to each RGB channel separately rather than the luminance channel, which seems like the more natural choice (perhaps with some associated adjustments to saturation). This has the effect that when bringing up shadows, parts of the image with tones near flatter parts of the curve get pushed strongly towards grey. Is there some way to avoid this issue and still use the tone curve? The only option I can see right now for bringing up shadows without this issue is the shadows slider in the basic tools box, which is substantially less flexible than the tone curve version.
   I can post sample images if it's not clear what I mean.

   Thanks,

Brian
   

[Schewe]

Is there some way to avoid this issue and still use the tone curve?

No...other than adjusting HSL (or Saturation) after a tone curve adjustment. Thomas Knoll made a conscience decision to tie saturation to the tone curve...adding contrast adds saturation, minus does the opposite. There is no luminance only method of adjusting curves in ACR/LR. You'll be better off learning to toolset and not fighting the behaviors. It is, what it is...

[bdosserman]

Hi,

No...other than adjusting HSL (or Saturation) after a tone curve adjustment. Thomas Knoll made a conscience decision to tie saturation to the tone curve...adding contrast adds saturation, minus does the opposite. There is no luminance only method of adjusting curves in ACR/LR. You'll be better off learning to toolset and not fighting the behaviors. It is, what it is...

Thanks for the reply. I don't completely understand your answer. Are you saying that the tone curve is actually making some more complicated calculation than what I was guessing, which involves both luminance and saturation? Or just that it is in fact applied to each RGB channel separately and that this has the effect of affecting saturation as well as luminance? If the latter, it seems to me it would even affect hue. Either way, do you know why Knoll decided to do it this way?

I'm attaching three images to illustrate my issue. (As it happens, this is the first photo I tried to process after buying LR yesterday). The first one is the original image (with a slight contrast/exposure adjustment). I was advised to try bringing the shadows up, so the second is obtained by sliding the shadows all the way up in basic tools. Looks good. But now, if I want to make finer adjustments and try to do the same thing in the tone curve, I reset the shadows slider in basic tools, and instead bring up the shadows slider in the tone curve. All the colors whose brightness falls into the flatter part of the curve get smashed to grey, making it very ugly. (As I'm looking at this, I think the curve version brought the shadows up a bit further, but the issue exists even if the slider hadn't been pushed as far)

Thanks again,

Brian

[Tim Lookingbill]

I am disappointed with the tone curve.

I'ld suggest you get to know all the tools available in LR and examine carefully what each does to your test image which BTW looks fine as is.

I'ld use the Shadow curve tool in the opposite direction "slid to the left" with the middle/shadow triangle adjusts slid far to the left to add definition between blackest black and the shadows. If it clips your black point in the histogram use the Shadow slider and/or Black Point adjust in Basic to lighten shadows.

What you're pointing out about the color behavior from one tool applied to one image doesn't point to an issue, but a lack of you not getting to know all the tools and how they work.

Play around and spend some time with the program. The things you'll discover on what you can do to an image using all the tools is where the fun's at.

[Schewe]

Are you saying that the tone curve is actually making some more complicated calculation than what I was guessing, which involves both luminance and saturation?

Yes, if you adjust the tone curve to reduce contrast, the saturation of colors are also reduced. If you increase the contrast, the saturation is increased. As to why Thomas did that, he thought it was more film-like. With film, increasing the contrast by push processing also increased the color saturation...pull processing reduced the contrast and reduced saturation.

As to your example...as you've found out, the Shadows slider in Basic behaves differently than the Shadows in the the Parametric Curves...rather than reseting the Basic Shadows, I would suggest keeping it where it was for the second version and adding a touch of positive adjustment in the Parametric Curves. It's also likely that you'll need to add a touch of increased saturation either in the Basic panel or the HSL panel. Tone mapping impacts color and saturation...adjusting color and saturation impacts the tone mapping. You can't do one without also adjusting the other.

[Tim Lookingbill]

Here's an example applying micro contrast through shadow tweaks I mentioned previously.

The one the left shows flat murky shadows on the leaf and dark areas within the petals just above the leaf.

The right shows what curve tweaks, Fill (Shadow Slider in Basic) and Parametric Shadow curves can do to maximizing shadow definition. I had to apply Hue tweaks to the purple channel to fix the red hue shift in the blue violet petals in the HSL panel.

Note the now saturated red hue that wasn't there in the triangular dark area within the cluster of the petals just above the leaf which brings out depth. If this dark area remained a neutral looking shadow it would've looked lifeless and ashen compared to the overall vibrance of the rest of the image.

I included the ACR screenshot of the shadow tweaks.

[bdosserman]

Yes, if you adjust the tone curve to reduce contrast, the saturation of colors are also reduced. If you increase the contrast, the saturation is increased. As to why Thomas did that, he thought it was more film-like. With film, increasing the contrast by push processing also increased the color saturation...pull processing reduced the contrast and reduced saturation.

I see, thanks for the explanation. I guess that makes sense, although I'd still prefer to have an option for the tonal curve to only adjustment luminance, keeping saturation closer to unchanged. Sort of like the current option to apply the curve to only one of the R, G, B channels.

As to your example...as you've found out, the Shadows slider in Basic behaves differently than the Shadows in the the Parametric Curves...rather than reseting the Basic Shadows, I would suggest keeping it where it was for the second version and adding a touch of positive adjustment in the Parametric Curves. It's also likely that you'll need to add a touch of increased saturation either in the Basic panel or the HSL panel. Tone mapping impacts color and saturation...adjusting color and saturation impacts the tone mapping. You can't do one without also adjusting the other.

OK, thanks for the suggestion.

Brian

[bdosserman]

Here's an example applying micro contrast through shadow tweaks I mentioned previously.

The one the left shows flat murky shadows on the leaf and dark areas within the petals just above the leaf.

The right shows what curve tweaks, Fill (Shadow Slider in Basic) and Parametric Shadow curves can do to maximizing shadow definition. I had to apply Hue tweaks to the purple channel to fix the red hue shift in the blue violet petals in the HSL panel.

Note the now saturated red hue that wasn't there in the triangular dark area within the cluster of the petals just above the leaf which brings out depth. If this dark area remained a neutral looking shadow it would've looked lifeless and ashen compared to the overall vibrance of the rest of the image.

I included the ACR screenshot of the shadow tweaks.

Thanks! These comments are very interesting. I'll definitely have to play around with all the various options to familiarize myself with what kind of effects can be achieved. I wouldn't have thought to actually try lowering the shadows slider on an image like this.

Brian

PS For the image I posted, the different versions were intended more to illustrate my initial question than anything else. I know I overdid bringing up the shadows, and was thinking of doing it in a more localized way. But your suggestions definitely open up new possibilities for me.

[stamper]

For me this link meant at last I fully understood all of what the sliders can achieve.

http://mulita.com/training/hns-r/

[Tim Lookingbill]

Thanks, stamper.

I've been looking for a tut like that for some time. That cleared things up on whether I should upgrade. I should've waited to buy a new computer and upgrading to CS6 instead of CS5 which doesn't have a lot of those slider features and certainly not PV2012.

Everything demonstrated in that video I've had to come up with workarounds mainly performed with the Point Curve in CS3 on over 1000 Raw images.

Nice to see the elegance in real time engineered into the newer slider driven interface.

Thanks for posting that Jardine video.

[stamper]

Imo a very good video and very well presented without any mistakes or repetition that is displayed in other videos. If I didn't have the LULA videos and three books then I would be tempted to buy the set. I searched for books that he might have written without success. I guess he isn't - yet - an author.

[bdosserman]

For me this link meant at last I fully understood all of what the sliders can achieve.

http://mulita.com/training/hns-r/

Finally got a chance to watch this. Thanks for the link! It was indeed quite informative.

Brian

[digitaldog]

As Jeff said, the product was designed this way. One can say it looks more film like, I'd ad it just looks better in most cases. And if you don't like the change in saturation, well there's a slider or two for that <g>.

Jardine's video's are IMHO must see TV. More than once! Very through, very well done.

[bdosserman]

As Jeff said, the product was designed this way. One can say it looks more film like, I'd ad it just looks better in most cases. And if you don't like the change in saturation, well there's a slider or two for that <g>.

Jardine's video's are IMHO must see TV. More than once! Very through, very well done.

Well, the problem is that the change in saturation isn't being applied evenly. It's specifically in the tones on the flatter part of the curve. So unless I'm missing some functionality, I don't know how the sliders would undo that. I see that saturation can be adjusted based on hue, but I don't see where I could adjust it based on brightness.

Brian

[digitaldog]

Well, the problem is that the change in saturation isn't being applied evenly.

It's a problem if you say it's a problem but I'd be more convinced if you showed us some examples of images where this is an issue and you have no way to get to the color appearance goal desired.

I suspect the reason the saturation isn't being applied 'evenly' is it was coded this way. Thomas has said in the past, it took a lot more engineering work to produce this effect then just applying NO saturation with the tone move. And considering what has to be tens of thousands if not millions of images that have been processed this way for a very long time, it's odd that only rarely do we hear a complaint about it and it usually surfaces after someone has read a rant about this behavior from Mr. Margulis or one of his followers.

Did you happen to also find this LuLa article:

http://www.luminous-landscape.com/essays/Curves.shtml

The last bee shot you uploaded (the one with the lower saturation) looks the least desirable to me. But it isn't my image so I defer here to your judgement.

[Peter_DL]

In RGB, any curve applied on all three channels also changes the R:G:B intensity ratios per pixel, by nature,
mainly effecting saturation and at times also the hue.

If we approximate and express a curve by a series of straight lines, any straight line with a negative offset increases saturation, any straight line with a positive offset decreases saturation.

The only 'curve' which does not do this is a straight line through the origin = linear scaling corresponding to exposure.

--

[bdosserman]

It's a problem if you say it's a problem but I'd be more convinced if you showed us some examples of images where this is an issue and you have no way to get to the color appearance goal desired.

I don't mean to blow things out of proportion. Now that I've had a chance to play with LR, I'm very happy with most if it. But I do think it would be an easier and more versatile tool for me personally if there were an option to have the tone curve leave saturation alone (or adjust it so that it is perceived as being left alone, since in my experience changing the L in HSL results in perceived differences to S, as well -- probably a defect of the HSL model). Would this be the sort of thing that could be addressed by programming a plug-in for LR?

I suspect the reason the saturation isn't being applied 'evenly' is it was coded this way. Thomas has said in the past, it took a lot more engineering work to produce this effect then just applying NO saturation with the tone move. And considering what has to be tens of thousands if not millions of images that have been processed this way for a very long time, it's odd that only rarely do we hear a complaint about it and it usually surfaces after someone has read a rant about this behavior from Mr. Margulis or one of his followers.

I've never heard of Margulis; I see that I've wandered blindly into what I can imagine has been a contentious discussion. My background is just that a) I'm a mathematician with a side interest in graphics, so I think about these things, and b) I've always been frustrated by the hue and saturation shifts in curves in the crappy photo-processing software I've used until now (I sometimes tried splitting to HSL and applying the curves to the L channel, but this usually didn't give great results because of what I mentioned above).

Did you happen to also find this LuLa article:

http://www.luminous-landscape.com/essays/Curves.shtml

Thanks for the link! I don't remember having read it when it was posted (and in any case if I had it wouldn't have meant nearly as much to me then as it does now). It's good to have the confirmation that at least LR is locking the hue when the tone curve is applied.

The last bee shot you uploaded (the one with the lower saturation) looks the least desirable to me. But it isn't my image so I defer here to your judgement.

Well, yes, that was my point. To me, the last image looks awful. But of course, I'm not saying one can't get decent results out of LR for this image. Based on the various suggestions and links people have posted, I've now found a couple different ways of doing it, but haven't yet settled on what I like best.

Brian

[bdosserman]

In RGB, any curve applied on all three channels also changes the R:G:B intensity ratios per pixel, by nature,
mainly effecting saturation and at times also the hue.

If we approximate and express a curve by a series of straight lines, any straight line with a negative offset increases saturation, any straight line with a positive offset decreases saturation.

The only 'curve' which does not do this is a straight line through the origin = linear scaling corresponding to exposure.

Yes, I realize this. But there's no reason software can't convert to HSL internally and apply the curve to L (or do something else more sophisticated) if one wanted to avoid this behavior.

Brian

[digitaldog]

Based on the various suggestions and links people have posted, I've now found a couple different ways of doing it, but haven't yet settled on what I like best.

There are a number of options to take that can result in a color appearance you desire, that's the good news.

Generally speaking, it's a good idea to work top down in Lightroom (I'd remove the word Generally if we could pick a camera profile up top). A lot of new LR users come to it from a Photoshop perspective so it's natural to gravitate to curves first. But I think if you see some of the suggested video's or play with the simple but powerful sliders above curves, you'll get the results faster. I find curves usage in LR to be very low which is the opposite of how I worked in Photoshop before I didn't need to use Photoshop for curves any more (meaning all this global tone and color work is done in Lightroom now). I find curves are useful in LR but for tiny tweaks.

But there's no reason software can't convert to HSL internally and apply the curve to L (or do something else more sophisticated) if one wanted to avoid this behavior.

I can't see a reason not to support such an option.

[Schewe]

In RGB, any curve applied on all three channels also changes the R:G:B intensity ratios per pixel, by nature,
mainly effecting saturation and at times also the hue.

But the Curve in LR/ACR is a hue protected curve...meaning it only impacts luminance and saturation (2/3 of the HSL potential). A normal tone curve in Photoshop does twist the hue, saturation and luminance. That's where the differences are. And again, Thomas did this on purpose. In my experience, when Thomas decides to something a certain way, he's usually right.