Lightroom Camera Profiles -- Discovered A Problem!

[ColSebastianMoran]

I have encountered a surprising problem with a Lightroom Camera Profile in which I get awful results, not just sub-optimal, but awful.

Here's the situation:
- A party, very much like a wedding reception
- Awful lighting, some fluorescent, some LED, some tungsten, low light
- Shot on E-M1 Mark II, a very current camera
- Shot at high ISO, eg 4000
- Processing in Lightroom, images require adjustments, some +1.5 stops, and a further +.5 on faces with the adjustment brush.

As expected, there's noise in the images. The default camera profile, "Adobe Standard" gives reasonable results in this extreme case.

So, here's the surprise: I would have thought the "Camera xxx" profiles would be robust, that is, give reasonable results even in extreme circumstances. When I try "Camera Muted" something awful happens on faces in some images: orange smears on the face and gritty black dots. Here's what I got with the "Camera Muted" camera profile.

Has anyone else experienced this kind of trouble with camera profiles and adjustment brush?

Camera Profile "Camera Muted," ISO 4000, Overall Exposure +1.45, Adjustment Brush on face +0.6, here's a 1:1 screen grab. I see two big problems here: 1) the orange smears, 2) the awful black grit.

 

[ColSebastianMoran]

For comparison, here's the same image area, at 1:1 using the "Adobe Standard" camera profile. Color is better, no orange smears.

But, there's still that black gritty noise?

ISO 4000, Profile "Adobe Standard," overall Exposure +1.45, Adjustment Brush on face +0.6.

 

[ColSebastianMoran]

How about the gritty black noise? Is there dust on the sensor (it would have to be very fine dust)? Is it an unavoidable by-product of the rather extreme adjustments? Or something else?

Adjustments in the image above include overall Exposure +1.45 and a further +.6 with the adjustment brush. Could it be the adjustment brush? For comparison, I'll just do overall Exposure +2.1.

The result is much better. An overall Exposure +2.2 is different from Overall +1.45 and brush +0.6. So the adjustment brush is creating the black grit, but why/how?

ISO 4000, Profile "Adobe Standard," Overall Exposure 2.2, no adjustment brush:

 

[ColSebastianMoran]

From above, we see that Overall Exposure +2.2 produces a far cleaner result vs Overall +1.45 and Adjustment Brush +0.6.

For comparison, let's try Adjustment Brush +2.2.

OK, now I've found the culprit! Adjustment Brush +2.2 is a gritty mess.

But, why?

Same image area, ISO 4000, Profile "Adobe Standard," Overall Exp +0.0, Adjustment Brush on face +2.1

 

[Dwig]

...

OK, now I've found the culprit! Adjustment Brush +2.2 is a gritty mess.
...

Check your Adjustment Brush settings. i've had some odd and very messy results when "Auto Mask" is checked.

 

[ColSebastianMoran]

Check your Adjustment Brush settings. i've had some odd and very messy results when "Auto Mask" is checked.

EDIT -- Double checking. Damn, triple THANKS Dwig. It WAS automask. Brushing w/ automask on the face produced a multitude of small unselected areas. Then, these are the source of the awful grit that gets worse the more I adjust with the brush.

Conclusions revised:
- Check for unexpected results with Camera Profiles. The orange smear
- Be careful with AutoMask, it can leave tiny areas unselected on any textured area (e.g. a face)
- The Lightroom Adjustment Brush is applied first, then global image adjustments, that can make a difference

 

[ColSebastianMoran]

A hint, a bit of internet reporting: "... global adjustments operate on top of the results of local adjustments (dependently). So once a brightening local adjustment is blowing out some highlight detail, your global adjustments in order to get to the desired overall tonality, won't then be able to pull those blown highlights back."

Sounds right, but this isn't the problem for the above example.

 

[dof]

Thanks to both Sebastian and DWIG for posting your findings here! Much appreciated.

 

[willie_901]

In-camera JPEG or raw?

 

[shawn]

Applying color profiles suggests it was RAW.

Shawn

 

[ColSebastianMoran]

Applying color profiles suggests it was RAW.

Shawn

Hi Shawn. Yes, these are RAW captures converted in Lightroom. I suspect the AutoMask problem could happen with jpg's as well.

 

[Godfrey]

The fact that you are having to add so much to exposure suggests to me that you're underexposing by quite a bit at the time of capture. Remember that dynamic range is reduced at high ISO settings, so the effective latitude of your exposure is smaller–you have to be closer to the mark in order that exposure adjustments not fall off the edge, and the edge is somewhat steeper than when shooting at more common ISO settings (ISO 100-1600).

On a pixel-by-pixel basis, when near the sensor's response edge, the results can be ambiguous and the result of applying local or global adjustments can go off the mark on a wildly divergent curve.

So what to be done: When shooting at elevated ISO settings (and on the smallish sensor of the Olympus E-M1 series cameras, for me that is anything over about ISO 1600) consider the scene carefully and give up trying to handle the detail in high contrast portions of the scene if you are trying to capture shadow values effectively. You don't have the dynamic range at those settings to stay happily away from the edge and layer base adjustments on top of local adjustments safely.

I have an E-M1 myself, and it is a marvelous camera. But I find its metering calibration when capturing at high ISO values tends to more underexposure than I feel is useful, so I adjust most exposures at ISO 3200 and beyond with a +0.3EV correction as a baseline. The resulting exposures, overall, require typically Lightroom adjustments in the -0.3 to +0.7 range, rather than in the +2.x range, and I do most of the correction using the Point Curve mode of the Tone Curve panel rather than the exposure sliders because I can tailor the curve with more accuracy and finer granularity than I can adjust the sliders.

I hope that helps ... 😀

G

 

[ColSebastianMoran]

The fact that you are having to add so much to exposure suggests to me that you're underexposing by quite a bit at the time of capture. Remember that dynamic range is reduced at high ISO settings, so the effective latitude of your exposure is smaller–you have to be closer to the mark in order that exposure adjustments not fall off the edge, and the edge is somewhat steeper than when shooting at more common ISO settings (ISO 100-1600). ... snip ...

G

Thanks, Godfrey. Yes, I agree the image I posted was certainly NOT optimal exposure.

Nonetheless, I was able in post to make usable images; the two failure areas were the "Camera Muted" profile (which I thought had potential) and the Auto Mask feature of the adjustment brush.

 

[ColSebastianMoran]

In-camera JPEG or raw?

No, we did not have in-camera jpg's in this shoot.

 

[Huss]

I'm curious, shooting at ISO 4000, then adding 1.5 EV, in essence is like shooting at ISO 10,000.
1. Do you get better results doing this than just shooting at 10,000? (I remember using an M9 it was better to add exposure post)
2. I also use an OMD (E10 Mk2) and really noticed that the m4/3 sensor is not as 'pliable' as larger ones.

 

[ColSebastianMoran]

I'm curious, shooting at ISO 4000, then adding 1.5 EV, in essence is like shooting at ISO 10,000.
1. Do you get better results doing this than just shooting at 10,000? (I remember using an M9 it was better to add exposure post)
2. I also use an OMD (E10 Mk2) and really noticed that the m4/3 sensor is not as 'pliable' as larger ones.

Huss, I have no conclusion on this.

A trusted source reported that the Sony mirrorless as of a couple years ago were "ISO-neutral" above ISO-800. That is, it didn't matter above ISO 800, adding ISO produced the same results as +exposure in post.

Adding ISO, you are trusting the camera engineers. Adding in post, you are trusting the software engineers (Adobe, etc.).

No comments on m4/3 being less or equally pliable. Except that the example face above was shot at ISO 4000 then plus 1.5 and another plus 0.5, so it's ISO 16000. I think the 1:1 above is pretty good for 16000. Good cameras in my view.

 

[willie_901]

Nothing Beats Signal-To-Noise Ratio

Nothing Beats Signal-To-Noise Ratio

I'm curious, shooting at ISO 4000, then adding 1.5 EV, in essence is like shooting at ISO 10,000.
1. Do you get better results doing this than just shooting at 10,000? (I remember using an M9 it was better to add exposure post)
2. I also use an OMD (E10 Mk2) and really noticed that the m4/3 sensor is not as 'pliable' as larger ones.

Yes.

Unless you have a camera with dual conversion-gain technology, the sensor sensitivity is constant. This is the camera' base or native ISO.

The camera ISO parameter (calibrated exposure index) and meter are used to generate an in-camera JPEG that will render with a useful brightness and not exceed the analog-to-digital converted maximum signal level. But the in-camera raw file is underexposed. In-camera the rendered brightness is achieved using electronic and, or digital signal amplification. A raw file will initially render with a useful brightness as well.

There are two types of underexposure. Here the world exposure refers to what happens when the shutter is open. An in-camera ISO above the base ISO modifies the data after the shutter closes

• Avoidable underexposure where ISO is simply set too high.
• Unavoidable underexposure where the appropriate shutter time and aperture result in sensor underexposure.

As Godfrey points out, the effects of avoidable underexposure are most obvious in shadow regions and for dynamic range. Highlight regions suffer too, but the effect is less obvious.

Underexposure always affects perceived image quality to some extent because underexposure decreases the data S/N.

I think the OP's problem is simply a S/N problem.

Some older cameras use analog signal amplification schemes (ISO settings) that actually reduce electronic noise levels. For instance, when underexposure is unavoidable, raw file rendered shadow regions will be more pleasing at ISO 800 than using base ISO and then increasing the brightness during post production.

Practically all newer cameras are psuedo-ISO invariant. This means the electric noise levels remain constant (within ~1/3 stop) as ISO increases. Now the shadow region rendering is nearly identical when an underexposed raw file is rendered brighter in post-production or rendered in camera (using ISO).

The "pliability" differences you observed are most likely due to differences in analog S/N. I observed this too as I upgraded cameras over the years.

Cameras with dual conversion-gain technology have two base ISOs. The lower ISO sensitivity provides the highest possible dynamic range and the higher ISO sensitivity provides the highest S/N. The former I used in bright light and the latter in lo light. Each of the two ranges are ISO invariant.

 

[ColSebastianMoran]

Willie, thanks for the info on sensors and ISO. There's a lot going on under the hood in today's cameras.

I think the OP's problem is simply a S/N problem.

I worried about that too, was it just high-ISO noise, working at the limits of the camera? The awful artifacts in the first image I posted in this thread were artificial, introduced by the profile and the mess created by Auto Mask on the adjustment brush.

See the third image which is relatively clean at, in effect, ISO 16000. Ever since the D200, and to some extent the D70, I have found great joy in making photographs indoors in the lighting we use everyday for social interactions. I'm shooting in conditions I was never able to make work with film.

 

[Huss]

Thank you Willie for the detailed explanation.

 

[willie_901]

...
I worried about that too, was it just high-ISO noise, working at the limits of the camera?

...

I think there are two separate issues at play.

Clearly there is an AutoMask issue.

Also, I think there are combinations of avoidable and unavoidable underexposure.

It's possible the latter makes the former more serious.

I prefer to use the phrase low S/N instead of high-ISO noise. This helps me think about exposure instead of noise.

As the S/N decreases some noise-signal components that would usually be below the analog-to-digital converter's detection threshold become digitized. An extreme example of this would be taking shots with the lens cap on in a dark room. The noise essentially becomes the signal. Typically color banding and other time-independent artifacts will appear when the S/N is very low. When you increase the Exposure slider in post-production, many time-independent artifacts will become visible. These rarely affect image quality because they are rarely digitized. It's possible some of the issues you see are due to this class of artifact.

As I mentioned above, if you know something about the ISO amplification scheme used in the camera you have some control over the electronic noise levels. But we have no control over the photon (a.ka. shot) noise. The problem isn't the noise, the problem is the low signal level.

The only solution is to use the lowest possible camera ISO setting that permits appropriate shutter and aperture settings. Now only camera and subject motion, along with DOF, limit exposure. This means the S/N for the data when the shutter is open will be maximized. FOR low ISO variant cameras some ISOs will be better than others. Also, in bright light you will be able to take full advantage of the camera's dynamic range. Maximizing exposure is useful even in very bright light.