Capture One Pro: recovering highlights and shadows not as good as ACR/LR

[Ligament]

I vastly prefer the colors and overall tones I can get from Capture one pro for all my cameras.

However, I tend to expose to the right with my cameras.

Traditionally, I've been very impressed with ACR/LR ability to recover highlights with the highlights slider and recover blacks with the blacks slider. In particular, the highlights recovery is amazing.

When developing the same image, I find much less highlight recovery capability in C1P. For example, in ACR/LR, I may need to bring the highlights slider 25% to the left to recover highlights, whereas in C1P I need to bring the highlight recovery tool 100% to the right to achieve similar recovery. However, I find that C1P at this setting (well, past the 20% setting) starts to squash not just my overexposed areas, but all my light tones and the image then looks flat.

I have similar issues with the blacks.

Is this an user issue (maybe I am missing something), or are the ACR/LR maths simply that much more advanced? Or, is ACR/LR lying to me about the recovery?

[ErikKaffehr]

Hi,

ACR used to be pretty good about highlight recovery. But the major benefit of LR is that both the highlights slider and the shadows sliders use content aware algorithms that are sort of HDR-mapping like tools playing tricks with local contrast enhancement.

Another thing is that Capture One has a quite aggressive 'film curve' that makes the images to bright, you could try the "linear" response instead.

With ACR/LR you could try different camera profiles. There used to be a few variants for the more popular cameras. You can also make your own camera profile.

Best regards
Erik

I vastly prefer the colors and overall tones I can get from Capture one pro for all my cameras.

However, I tend to expose to the right with my cameras.

Traditionally, I've been very impressed with ACR/LR ability to recover highlights with the highlights slider and recover blacks with the blacks slider. In particular, the highlights recovery is amazing.

When developing the same image, I find much less highlight recovery capability in C1P. For example, in ACR/LR, I may need to bring the highlights slider 25% to the left to recover highlights, whereas in C1P I need to bring the highlight recovery tool 100% to the right to achieve similar recovery. However, I find that C1P at this setting (well, past the 20% setting) starts to squash not just my overexposed areas, but all my light tones and the image then looks flat.

I have similar issues with the blacks.

Is this an user issue (maybe I am missing something), or are the ACR/LR maths simply that much more advanced? Or, is ACR/LR lying to me about the recovery?

[Paul2660]

You can also try a local adjustment layer on the areas with the highlights.

I personally don't like to use the highlights slider on a 100 percent use as it can have possible adverse effects to other parts of the file.

As Eric mentioned  the linear curve at times will also help to tone down the extreme push C1 makes to files that may blow out a highlight.

Paul C

[ErikKaffehr]

Hi,

in general, I would say that it is a very good idea to frequently check out your exposures with RawDigger: http://www.rawdigger.com

This tool shows the actual raw histograms and tells a lot about the raw exposures.

Best regards
Erik

[Bart_van_der_Wolf]

I vastly prefer the colors and overall tones I can get from Capture one pro for all my cameras.

However, I tend to expose to the right with my cameras.

Traditionally, I've been very impressed with ACR/LR ability to recover highlights with the highlights slider and recover blacks with the blacks slider. In particular, the highlights recovery is amazing.

Hi,

LR offers very good highlight reconstruction in Process 2012. Do note, that's not the same as recovery. One can only recover something that isn't totally lost. It sounds like you are exposing too far to the right, and depend on LR to make something presentable from that. LR reconstructs totally (in 3 channels) blow highlights as white, instead of gray, no matter the amount of exposure correction or highlight corrections. That's reconstruction.

When developing the same image, I find much less highlight recovery capability in C1P. For example, in ACR/LR, I may need to bring the highlights slider 25% to the left to recover highlights, whereas in C1P I need to bring the highlight recovery tool 100% to the right to achieve similar recovery. However, I find that C1P at this setting (well, past the 20% setting) starts to squash not just my overexposed areas, but all my light tones and the image then looks flat.

As already mentioned by others, make sure to start with a linear tonecurve in Capture One. The (default) film tonecurve will seemingly blow out up to almost a stop of highlights if you exposed the Raw data to the right (because Phase One backs traditionally underexpose to protect the highlights). You can use an application like Fast Raw Viewer or RawDigger to evaluate the actual Raw exposure level. With a linear tonecurve, you also retain better contrast in the highlights you might have to the recover later. Make sure to not clip more than 1 channel (or 2 at most) if you want realistic recovery potential. Best is not to clip to begin with.

I have similar issues with the blacks.

Hard to say how to handle that. First, don't worry about some clipped black, second, don't try to make shadows too light. Work with the Contrast control and Exposure control, if the scene contrast is really that high. Contrast will allow to fit the scene range in the histogram with minimal clipping, and then, only then, does one use the other controls to tonemap the entire range to a pleasing contrast where it matters.

Is this an user issue (maybe I am missing something), or are the ACR/LR maths simply that much more advanced? Or, is ACR/LR lying to me about the recovery?

It's a mix. LR is pretty good at an overall tonemapping of difficult scenes, but in doing so you risk losing sparkling highlights in an attempt to fit in everything else. With a negative 100 Highlights that can be remedied, but that leaves a lot less wiggleroom for the rest.

Capture One is not as flexible, possibly with more properly exposed images in mind (using fill light or diffusors to prevent issues in advance), but is has become better over the last few release versions. If I have 'impossible' scenes, I use a dedicated tonemapping application (which dwarfs LR's capabilities), but for normal challenging scenes one or more Adjustment layers can locally tame extreme highlights and locally brighten shadows in several ways.

Cheers,
Bart

[Ligament]

Hi,

LR offers very good highlight reconstruction in Process 2012. Do note, that's not the same as recovery. One can only recover something that isn't totally lost. It sounds like you are exposing too far to the right, and depend on LR to make something presentable from that. LR reconstructs totally (in 3 channels) blow highlights as white, instead of gray, no matter the amount of exposure correction or highlight corrections. That's reconstruction.

As already mentioned by others, make sure to start with a linear tonecurve in Capture One. The (default) film tonecurve will seemingly blow out up to almost a stop of highlights if you exposed the Raw data to the right (because Phase One backs traditionally underexpose to protect the highlights). You can use an application like Fast Raw Viewer or RawDigger to evaluate the actual Raw exposure level. With a linear tonecurve, you also retain better contrast in the highlights you might have to the recover later. Make sure to not clip more than 1 channel (or 2 at most) if you want realistic recovery potential. Best is not to clip to begin with.

Hard to say how to handle that. First, don't worry about some clipped black, second, don't try to make shadows too light. Work with the Contrast control and Exposure control, if the scene contrast is really that high. Contrast will allow to fit the scene range in the histogram with minimal clipping, and then, only then, does one use the other controls to tonemap the entire range to a pleasing contrast where it matters.

It's a mix. LR is pretty good at an overall tonemapping of difficult scenes, but in doing so you risk losing sparkling highlights in an attempt to fit in everything else. With a negative 100 Highlights that can be remedied, but that leaves a lot less wiggleroom for the rest.

Capture One is not as flexible, possibly with more properly exposed images in mind (using fill light or diffusors to prevent issues in advance), but is has become better over the last few release versions. If I have 'impossible' scenes, I use a dedicated tonemapping application (which dwarfs LR's capabilities), but for normal challenging scenes one or more Adjustment layers can locally tame extreme highlights and locally brighten shadows in several ways.

Cheers,
Bart

Wow, fantastic advise thank you Bart! I was not aware of the difference between recovery and reconstruction with the highlights slider.

Attached is a RawDigger histogram on a fairly typical image which I ETTR, and one with which I am having difficultly reconstructing highlights in with C1P vs. ACR/LR...

[ErikKaffehr]

Hi,

I would not argue with Bart, as he is usually right :-)

The tools in Lightroom are a bit limited and they can introduce artefacts. That said, in many cases LR gives a well working single step solution.

I was looking a lot into HDR mapping solutions before process variant 2012 in Lightroom. With PV 2012 LR gave me solutions to most problems, in a simple parametric workflow that did not force me in working with files rendered in TIFF.

For me the balance is in favour of staying with parametric workflow inside Lightroom. My approach is to make as much work as possible in a parametric workflow but going into Photoshop when parametric workflow is no longer applicable.

Best regards
Erik

Hi,

...
Capture One is not as flexible, possibly with more properly exposed images in mind (using fill light or diffusors to prevent issues in advance), but is has become better over the last few release versions. If I have 'impossible' scenes, I use a dedicated tonemapping application (which dwarfs LR's capabilities), but for normal challenging scenes one or more Adjustment layers can locally tame extreme highlights and locally brighten shadows in several ways.

Cheers,
Bart

[Ligament]

Hi,

I would not argue with Bart, as he is usually right :-)

The tools in Lightroom are a bit limited and they can introduce artefacts. That said, in many cases LR gives a well working single step solution.

I was looking a lot into HDR mapping solutions before process variant 2012 in Lightroom. With PV 2012 LR gave me solutions to most problems, in a simple parametric workflow that did not force me in working with files rendered in TIFF.

For me the balance is in favour of staying with parametric workflow inside Lightroom. My approach is to make as much work as possible in a parametric workflow but going into Photoshop when parametric workflow is no longer applicable.

Best regards
Erik

Definitely not arguing with Bart :-)

[Bart_van_der_Wolf]

Attached is a RawDigger histogram on a fairly typical image which I ETTR, and one with which I am having difficultly reconstructing highlights in with C1P vs. ACR/LR...

Great, that allows to get a bit more specific, instead of having to talk in generalities.

I know that specific image content will determine the specific shape of the Raw histogram, but there are several additional things to be aware of.

Depending on the Raw format, but usually the case, the Raw histogram shows the situation before White Balancing. That means that when a Raw converter applies WB, some channels could (relatively to the others) become even more shifted to the right. If the Raw converter doesn't allow to manage that ourselves somehow, we are at the mercy of how the Raw converter deals with that.

RawTherapee for example, allows to compensate for the White point being dominated by a single channel (after White Balancing) by proportionally reducing the other channels. But that does assume relatively little clipping of any channel, for the best results.

In your RawDigger example, as I expected a bit, we can see 3 channels (1 only barely) clipped to the maximum in Raw, and that's even before White Balancing. So that will be a challenge. LR will do a decent job of reconstructing the missing parts, but they are still guesses (some of them are good, but not perfect). Capture One (especially with a 'film' tone curve) will only be confused more. To have at least a fighting chance, a linear curve in this and similar cases is a must if you do not want to lose more of what is available.

In C1, you will still be faced with the clipped/missing Raw data, which will be handled differently. What I would try in this particular case, is to work with possibly the Levels tool by trying to recreate a proper White Balance, by reducing the relative output levels (at the top of the dialog) of the least clipped channels. The most clipped channel should clip at maximum, the others at a lower level. That may produce funny colors at the clipped maximum, but restore color balance at lower non-clipped levels. You can clip the 'funny' highlight colors to White afterwards with another control, like Curves, the same for all three channels.

It's a bit hard to explain, but I hope you get the drift.

Cheers,
Bart

[Ligament]

Great, that allows to get a bit more specific, instead of having to talk in generalities.

I know that specific image content will determine the specific shape of the Raw histogram, but there are several additional things to be aware of.

Depending on the Raw format, but usually the case, the Raw histogram shows the situation before White Balancing. That means that when a Raw converter applies WB, some channels could (relatively to the others) become even more shifted to the right. If the Raw converter doesn't allow to manage that ourselves somehow, we are at the mercy of how the Raw converter deals with that.

RawTherapee for example, allows to compensate for the White point being dominated by a single channel (after White Balancing) by proportionally reducing the other channels. But that does assume relatively little clipping of any channel, for the best results.

In your RawDigger example, as I expected a bit, we can see 3 channels (1 only barely) clipped to the maximum in Raw, and that's even before White Balancing. So that will be a challenge. LR will do a decent job of reconstructing the missing parts, but they are still guesses (some of them are good, but not perfect). Capture One (especially with a 'film' tone curve) will only be confused more. To have at least a fighting chance, a linear curve in this and similar cases is a must if you do not want to lose more of what is available.

In C1, you will still be faced with the clipped/missing Raw data, which will be handled differently. What I would try in this particular case, is to work with possibly the Levels tool by trying to recreate a proper White Balance, by reducing the relative output levels (at the top of the dialog) of the least clipped channels. The most clipped channel should clip at maximum, the others at a lower level. That may produce funny colors at the clipped maximum, but restore color balance at lower non-clipped levels. You can clip the 'funny' highlight colors to White afterwards with another control, like Curves, the same for all three channels.

It's a bit hard to explain, but I hope you get the drift.

Cheers,
Bart

I believe I get it. Bart, just fantastic. Thank you for your time.

BTW, this particular image was shot in color (obviously) but I am making it into a B&W image. Does that change your advise?

[ErikKaffehr]

Hi Bart,

Note that logarithms are used on the Y-axis, the number clipped pixels is quite small around 47k and below 1%. So it looks like a resonable ETTR exposure.

On the other hand the green and blue channels don't roll down to zero, so clipping is probably not limited to specular highlights. Another point may be that the raw data, as plotted, is in 12 bits representation. I would assume that C1 works with 16 bits internally, so it would be able to shift data for white balance quite a lot (16X times) without inducing additional clipping.

What is your take on just using an exposure bias?

As far as I can see, there should not be any real issues with handling the highlights as less than 1% of the pixels are beyond recovery. But I guess that it can be hard to have a good tonal separation in the bright areas. Using a (more) linear curve helps with that as the derivatives in the highlights would be larger with linear response. The downside may be that C1 colour profiles are intended to be used with film like curves.

The highlights slider in LR/ACR probably uses some local contrast adoption, while the HDR sliders in C1 seem to do more of highlight compression, although this has been improved in V9 from V8.

Best regards
Erik

Great, that allows to get a bit more specific, instead of having to talk in generalities.

I know that specific image content will determine the specific shape of the Raw histogram, but there are several additional things to be aware of.

Depending on the Raw format, but usually the case, the Raw histogram shows the situation before White Balancing. That means that when a Raw converter applies WB, some channels could (relatively to the others) become even more shifted to the right. If the Raw converter doesn't allow to manage that ourselves somehow, we are at the mercy of how the Raw converter deals with that.

RawTherapee for example, allows to compensate for the White point being dominated by a single channel (after White Balancing) by proportionally reducing the other channels. But that does assume relatively little clipping of any channel, for the best results.

In your RawDigger example, as I expected a bit, we can see 3 channels (1 only barely) clipped to the maximum in Raw, and that's even before White Balancing. So that will be a challenge. LR will do a decent job of reconstructing the missing parts, but they are still guesses (some of them are good, but not perfect). Capture One (especially with a 'film' tone curve) will only be confused more. To have at least a fighting chance, a linear curve in this and similar cases is a must if you do not want to lose more of what is available.

In C1, you will still be faced with the clipped/missing Raw data, which will be handled differently. What I would try in this particular case, is to work with possibly the Levels tool by trying to recreate a proper White Balance, by reducing the relative output levels (at the top of the dialog) of the least clipped channels. The most clipped channel should clip at maximum, the others at a lower level. That may produce funny colors at the clipped maximum, but restore color balance at lower non-clipped levels. You can clip the 'funny' highlight colors to White afterwards with another control, like Curves, the same for all three channels.

It's a bit hard to explain, but I hope you get the drift.

Cheers,
Bart

[Ligament]

Here is the histogram in linear mode

[ErikKaffehr]

Hi,

That histogram seems quite OK to me!

This posting by the "Image Quality Professor" may offer some insight:

http://blog.phaseone.com/make-images-stand-clarity/

Best regards
Erik

Here is the histogram in linear mode

[Chairman Bill]

Capture One is lacking in a number of areas, and even Aperture (which hasn't had, and won't get, an update in a very long time) still outperforms it in many areas - actually most if not all of them, apart from raw conversion.

[Bart_van_der_Wolf]

Here is the histogram in linear mode

Yes, that looks less dramatic than the Logarithmic Histogram counts, but the fact remains that the Blue and both Green planes have a fair number of clipped pixels, and Red a few. I have no Idea if those clipped pixels are the ones that cause significant visual effects and how they impact Whitebalancing and the color of (specular) highlights.

So in this particular case, it looks like you have to choose to clip some of the highlights instead of trying to recover them, they are blown out so recovering in not going to be that succesful. Depending on the image, it's best to focus on the tonality of the rest of the image first, then use an adjustment layer to locally see how much of the clipping can be recovered (maybe the Red clipping is in pixels that are not also Green clipped at the same time, a histogram won't tell that).

I'm involved in a project that deals with shooting a lot of products which show specular highlights, like silverware and glossy ceramics (e.g. vases), and glass objects. If I were to do a straight highlight preserving Raw conversion in C1, then the images would look dark and have mediocre tonality, unless I spent an inordinate amount of time on Curves. So I do a normal conversion based on the main object(s), and then add a local adjustment to mask in the highlights if possible. And it usually is posible, e.g. with exposure at -0.9 and Highlight HDR of 10, and a gradually built up mask with a low flow amount, depending on the amount of output clipping I want to preserve).

Do note that not all highlights need to always render as totally unclipped. Specular highlights are a reflection of an illuminant/lightsource, often with focused intensity due to a convex surface, and there is often not much sense in trying to create color in those very bright highlights if those highlights show a mirror image of the White Balanced main illuminant without added reflected surface color.

Cheers,
Bart