Taming Blown Highlights

[NigelG]

Advertisement (gone after registration)

6 hours ago, adan said:

Expose for the highlights - expose for the highlights - expose for the highlights!

Then "tame" the shadows.

The M10 has enormous extra DR in the shadows. Most of the reviewers remarked on it in 2017.

Leica hid that with a contrasty tone curve to get back some of the punch and saturation of the M9 CCD, straight out of the camera.

But it is easy to counteract with the exposure/shadows/contrast/blacks sliders. Don't be afraid to push the shadows to +100 if needed - the M10 also has very good chroma/color invariance (hues stay the same even when "pushed" or opened up 2-4 stops).

Here's an example from my first summer with the M10 (2017), once I figured out what was going on. What came out of the camera - and then what was really there. Image shot at minus -0.7 Exposure comp. Nothing is blown except a couple of pin-point specular reflections of the sun in the man's water-bottle and watch bezel.

It may even look overcooked and  "too HDR" to some - it does to me - but I'm just showing how much is available in the M10 shadow data. You only have to use as much as you want.

A little bump to the mid-tone contrast with curves can tame the HDR appearance.

Straight from the camera (90mm Summarit f/2.5):

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

"Tamed"/recovered shadows.

 

 

I think you’re right re the overcooking of the second image.

I wasn’t there and didn’t see the “shot“ but I think the first version, maybe actually with no highlight-protecting-dialled-in exposure compensation, might have been on the money “in camera”.
 

To my eye a slight uplift in the exposure of the first version would probably be my “memory” of the moment…figures in partial shadow and quite harsh yet maybe starting-to-lengthen sunlight.

I don’t think I would necessarily have “seen” the second image as presented, and don’t really know why it would be considered somehow “better” on the basis that the shadow detail has been brought out in PP…but maybe that’s just me.🤔

 

[adan]

Let's say the metering gives you ISO 1600, f/5.6, 1/250 sec.

OK

Let's say you want to preserve highlights by reducing exposure by one stop and shoot with 1/500 sec instead.

Correct - you will be giving the sensor an "ISO 3200" exposure (manually, or using exposure compensation of -1 = same thing), while the camera is set to 1600. Personally, I use exp. comp of -0.67 rather than -1.00, most of the time, and find that does the job.

Unfortunately, that is a bad idea, since it also increases noise.

Well, both "bad idea" and "noise" are subjective abstract values. I've never had a picture rejected for "noise." Not by publications, nor by galleries, nor in independent print sales. It is just like with film. If the situation allows for Kodachrome 25, that's great - and if the situation calls for Fujichrome 400 pushed two stops, then that is what one uses, and doesn't expect Kodachrome quality.

A better approach is to shoot at ISO 800 instead of ISO 1600 (f/5.6, 1/250 sec).

Well again, "better" is subjective. That is not a better technique if I'm trying to stop the action of moving subjects in dim light - which is when I use higher ISOs.

The noise in the image, which is determined by the exposure (not ISO), stays the same. However, the image needs to be lifted by one stop in the post.

A picture shot at 1600 and normal exposure, and an image shot at ISO 800 and "lifted" one stop amount to the same thing - you are effectively right back to your original ISO, with identical S/N ratio and identical noise. What I showed in the prevous samples was exactly and precisely "lifting the image in post."

With an invariant sensor like the M10's, it doesn't matter whether you lift the image, or the camera does by giving "normal" exposure in the first place.

This works better with raw images than with JPGs, though. Of course.

.............

I'll go you 2 stops better. Here's an M10 image at ISO 6400, exposed -0.67 stops to hold the highlights (i.e. given the light for ISO 10000). The only thing slightly blown is the blue spotlight lower right - and only in the blue channel (it is still blue, not gray). And how it looked with the shadows "lifted."

Processed about like the previous example, except less extremely, and with appropriate noise-reduction for the situation.

Straight from camera, and then adjusted. Click through for larger versions.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Herr Barnack]

9 hours ago, SrMi said:

What do you mean by "correct meter reading"?

"Correct" according to the camera's meter when exposure compensation is set to 0.0 - use that as a starting point, shoot in manual mode and reduce exposure by -1EV to -2EV.  Make more than one shot at different exposures so you have options to work with in post.

Edited June 23, 2022 by Herr Barnack

[SrMi]

3 minutes ago, adan said:

Let's say you want to preserve highlights by reducing exposure by one stop and shoot with 1/500 sec instead.

Correct - you will be giving the sensor an "ISO 3200" exposure (manually, or using exposure compensation of -1 = same thing), while the camera is set to 1600. Personally, I use exp. comp of -0.67 rather than -1.00, most of the time, and find that does the job.

Calling it an  "ISO 3200" exposure is confusing, as ISO has no part in an exposure. I would call it "ISO 3200" metering.

In my "bad" example, you would be shooting ISO 3200, f/5.6, 1/500 (-1EV) instead of ISO 3200, f/5.6, 1/250 (0EV). My suggestion is to keep the shutter speed and lower ISO instead.

11 minutes ago, adan said:

Unfortunately, that is a bad idea, since it also increases noise.

Well, both "bad idea" and "noise" are subjective abstract values. I've never had a picture rejected for "noise." Not by publications, nor by galleries, nor in independent print sales. It is just like with film. If the situation allows for Kodachrome 25, that's great - and if the situation calls for Fujichrome 400 pushed two stops, then that is what one uses, and doesn't expect Kodachrome quality.

I should have used the term "wasteful" instead of "bad." Similarly, why shoot ISO 6400, f/5.6, and f/1000 when the shutter speed would be OK to use ISO 1600, f/5.6, 1/250? While both settings may generate an excellent image, the first one is wasteful and adds unnecessary noise.

I am not saying that I am always following my suggestions, though, but I try :).

18 minutes ago, adan said:

A better approach is to shoot at ISO 800 instead of ISO 1600 (f/5.6, 1/250 sec).

Well again, "better" is subjective. That is not a better technique if I'm trying to stop the action of moving subjects in dim light - which is when I use higher ISOs.

In my example, 1/250 was desired shutter speed and was increased to 1/500 only to apply -1EV, but not to stop the action.

20 minutes ago, adan said:

The noise in the image, which is determined by the exposure (not ISO), stays the same. However, the image needs to be lifted by one stop in the post.

A picture shot at 1600 and normal exposure, and an image shot at ISO 800 and "lifted" one stop amount to the same thing - you are effectively right back to your original ISO, with identical S/N ratio and identical noise. What I showed in the prevous samples was exactly and precisely "lifting the image in post."

With an invariant sensor like the M10's, it doesn't matter whether you lift the image, or the camera does by giving "normal" exposure in the first place.

I assume you mean 1600 normal metering (0EV). To clarify: the SNR is determined by exposure (shutter speed and aperture), not by ISO. So with the same aperture and shutter speed, you have (almost) the same SNR at ISO 800 and ISO 1600, but with ISO 800, you have more highlight protection.

M10 sensor is invariant only at ISO 400 and above.

[SrMi]

16 minutes ago, Herr Barnack said:

"Correct" according to the camera's meter when exposure compensation is set to 0.0 - use that as a starting point, shoot in manual mode and reduce exposure by -1EV to -2EV.  Make more than one shot at different exposures so you have options to work with in post.

Spot, center-weighted, matrix metering, and different framing give you all different metering results. I could describe optimal exposure, but would not know how to describe correct exposure.

[Herr Barnack]

1 hour ago, SrMi said:

Spot, center-weighted, matrix metering, and different framing give you all different metering results. I could describe optimal exposure, but would not know how to describe correct exposure.

Optimal is a more accurate description of what I was attempting to express than correct.  So let's use the term optimal.

It seems to me that for evaluating optimal exposure for a given scene, matrix metering would be the optimal metering mode.  Of course, every scene is different so using good judgement enters in to the equation.

As for framing, I would decide on my framing and then do my metering and figure out if -1EV (or more) manual compensation is applicable, or if it is even necessary to avoid blowing out the highlights.

[lct]

Advertisement (gone after registration)

7 hours ago, SrMi said:

I assume you mean 1600 normal metering (0EV). To clarify: the SNR is determined by exposure (shutter speed and aperture), not by ISO. So with the same aperture and shutter speed, you have (almost) the same SNR at ISO 800 and ISO 1600, but with ISO 800, you have more highlight protection.

+1 but i don't understand how this better highlight protection is working. Would you mind to elaborate on this? Just curious.

[Adam Bonn]

Digital ISO is a fallacy. It's effectively the same as turning up the knob on a Hi-Fi amplifier the music (picture) gets louder (brighter) at the cost of sound quality (noise/DR)

When you make an amplification in camera (ie ISO dial) you're amplifying all the parts of the picture, when you chose SS/A values but with a lower than camera recommended ISO value, you have the option in post of selectively brightening only the dark &/or mid tones, which is a way that many people like to work (which is not what I'm showing below)

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

(zero edits)

 

+3 stops on the exposure slider on xx14.DNG

There is no increase of DR or in noise reduction or in highlight retention by "pushing ISO-less"* sensors like this, merely a way to circumnavigate nr tag choices and amplification decisions that the camera makes by handling these choices oneself. 

As in many things, all roads lead to Rome centurion 

---------

*older cameras with pre-ADC amplification and modern camera with dual (sic) ISO ranges need to play this game a little differently

------------

Aren't forums wonderful... OP = so guys, highlights at low ISO? Twenty-odd posts later forum =  at high ISO shoot like this 😅

 

[lct]

Thank you @Adam Bonn  but i still don't understand how clipping can be avoided this way. I mean if it's clipped at f/2.8, 250/s, 3200iso the light does not change when i set iso to 1250. Exposure remains the same in the narrow sense our colleagues are using here does it not? 

[Adam Bonn]

38 minutes ago, lct said:

Thank you @Adam Bonn  but i still don't understand how clipping can be avoided this way. I mean if it's clipped at f/2.8, 250/s, 3200iso the light does not change when i set iso to 1250. Exposure remains the same in the narrow sense our colleagues are using here does it not? 

It can't, if it clips it's gone.

By pushing the ISO after the shot in post, we're pushing the brightness up to where we want it, rather than bringing it down from where perhaps we don't, plus we're not letting are raw app automatically apply xyz amount of NR based on the ISO (which again we could also turn down ourselves)

For me (OMWV) it's a bit like the profiles I make, far easier to add saturation in later during post, either globally or on a colour range, than take it out later. Equally I prefer to try and not clip highlights that I want, then after the fact add brightness where I want it. (or even, heaven forbid, have shots that need very little doing to them )

But yeah 100% correct if ISO800 F2 and 1/125 clips, then so does 1/125 F2  and ISO200

Back to the OP (oh yeah that!!)

Stop down, buy an ND filter.

I'd be curious to know where @erniethemilk is having a problem?

Shots where the SS is 3-4000 (not really a lot of latitude for metering error) or more in the (Leica) midrange of 750-1500, in which case the question is, why wasn't a higher SS used, ie what made you/the camera pick that speed?

There might be a degree of workflow involved too... IME (FWIW) the M10 isn't a camera that you just drag the shadow slider up in post there's a juggling act of using the exposure/shadow/highlight sliders to get the desired image data right, then using the blk/wht/contrast sliders to get the feel of the photo that you want.

But y'know... workflow's are like roads to Rome  

[lct]

16 minutes ago, Adam Bonn said:

It can't, if it clips it's gone.

How do you explain my pics above then?

 

[SrMi]

1 hour ago, lct said:

Thank you @Adam Bonn  but i still don't understand how clipping can be avoided this way. I mean if it's clipped at f/2.8, 250/s, 3200iso the light does not change when i set iso to 1250. Exposure remains the same in the narrow sense our colleagues are using here does it not? 

Note that ISO does not change the sensitivity of the sensors. 
Assuming there is not much light to use base ISO, no sensor pixel will be fully saturated (clipped). However, clipping will occur because the pixel value was "amplified" by a higher ISO. The clipping can be avoided with less "amplification" (lower ISO).

[SrMi]

20 minutes ago, Adam Bonn said:

But yeah 100% correct if ISO800 F2 and 1/125 clips, then so does 1/125 F2  and ISO200

That is not necessarily true. For example, if clipping occurs because of ISO "amplification," then clipping can be avoided by using less "amplification."

[evikne]

Even though the M10 can recover a little more than 1 EV from the highlights at ISO 200, I quickly learned to never "expose to the right", like I always did with my Canon camera. A picture that has been adjusted 1 EV down to the desired exposure, look very different from one that has been adjusted 1 EV up to the desired exposure. The latter has a much richer and more natural tonality. And an underexposed image can be lifted surprisingly high in LR without losing detail in the highlights.

[SrMi]

2 minutes ago, evikne said:

Even though the M10 can recover a little more than 1 EV from the highlights at ISO 200, I quickly learned to never "expose to the right", like I always did with my Canon camera. A picture that has been adjusted 1 EV down to the desired exposure, look very different from one that has been adjusted 1 EV up to the desired exposure. The latter has a much richer and more natural tonality. And an underexposed image can be lifted surprisingly high in LR without losing detail in the highlights.

AFAIK, all sensors in modern cameras operate only in their linear range. That means that, except for noise, the colors should be the same and depend on the post-processing. So far I have not seen an example where correct ETTR (no clipping) produces different or worse colors. 
I would like to see M10 examples where that is not the case. 

The issue with lifting an underexposed image is the noise in the shadow, not the highlight clipping.

[Adam Bonn]

1 hour ago, SrMi said:

That is not necessarily true. For example, if clipping occurs because of ISO "amplification," then clipping can be avoided by using less "amplification."

I think maybe there's some confusion here by the definition of clipping.

For me clipped means, gone, blown, cut, missing

If the data isn't clipped, but appears too bright, then reducing brightness (highlight slider/expo slider) will restore it.

If it's clipped, it's gone, outta here, Elvis has left the building.

[Adam Bonn]

15 hours ago, lct said:

Blown highlights can be avoided by reducing iso values w/o changing aperture or shutter speeds (Fuji X-E2, Summarit 50/2.5, 1/125s, f/2.8, 3200 & 1250 iso).

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

I think that you could reach the look of the second image (from the first) using highlight recovery. But it's not my shot to know.

Open both files in RAW Digger and it'll indicate any clipping

[Adam Bonn]

If we shoot a contrasty image exposing to preserve HL details, then in post lift the shadows two stops we have effectively made an image that has ISO 200 bright bits and ISO 800 for the shadows.

Whatever we chose to do we have X amount of DR to play with. If we exceed 'X' with our exposure choices then that data is gone.

We can go dark then make brighter, or go brighter then make darker, but only within X

The reason that some cameras (listed as having similar DR) behave very differently with exposure latitude is to do with stuff I don't really understand (!!) -but many on here do- but basically things like how the camera engine creates the RAW (in-built tonality, eg M10 default contrasty tone curve) and also what state of amplification the sensor is in before we do anything to make a picture with it. The same base chip can be coded to perform at a variety of base ISOs by the camera engineers. My completely uneducated hunch with the M10 is that it's a similar chip to the 240 (and same CMOSIS OEM no?) but with better ISO, so I assume it's in a reasonably stretched (sic) state to achieve higher ISO than the 240 (as well as Leica's not quite the truth base ISO claim), ergo the M10 lends it's self well to shadow recovery, but not highlights, because it's always running hot (sic).

But just a hunch.

[SrMi]

23 minutes ago, Adam Bonn said:

I think maybe there's some confusion here by the definition of clipping.

For me clipped means, gone, blown, cut, missing

If the data isn't clipped, but appears too bright, then reducing brightness (highlight slider/expo slider) will restore it.

If it's clipped, it's gone, outta here, Elvis has left the building.

There is clipping on the sensor, and there is clipping in the DNG file.
Without clipping on the sensor, you could avoid clipping in the DNG by using an ISO low enough.
With clipping on the sensor, there will be clipping in the DNG even when shooting at base ISO. In that case, the clipping can be avoided only by reducing the exposure (aperture, shutter speed).
When you see clipping in the DNG at higher ISOs, it does not mean that the sensels were saturated, but maybe the ISO "amplification" clipped the data.

[Advertisement]

Hallo,

Look here: Leica M

[adan]

1 hour ago, Adam Bonn said:

But y'know... workflow's are like roads to Rome  

Exactly.

1 hour ago, Adam Bonn said:

IME (FWIW) the M10 isn't a camera that you just drag the shadow slider up in post there's a juggling act of using the exposure/shadow/highlight sliders to get the desired image data right, then using the blk/wht/contrast sliders to get the feel of the photo that you want.

Also exactly, which is why in my first post I mentioned all those sliders.

-------------

With negative film, there is a basic exposure mantra: "Expose for the shadows, develop for the highlights." Expanded to its most complex iteration in Ansel Adams' "Zone System" and variations.

With direct-to-positive processes like digital, that gets reversed: "Expose for the highlights, develop (or post-process) for the shadows."

In either case, the goal is to avoid "blank" parts of the picture - detail-free film shadows, or blown digital highlights.