Exposing to the Right & PP

[Glenn NK]

The sunrise image simply has too much dynamic range for ANY method of exposure - this has absolutely nothing to do with ETTR or any other method of exposure. There isn't a sensor (yet) that can handle the range in this image.

This scene simply requires at least two exposures blended together - one to just avoid clipping the highlights of the sky, and one several stops darker to capture the shadows - with the two images combined into a narrower dynamic range.

Neither the computer screen nor a printer can handle the total DR of this scene - it needs to be compressed or the compromises will be difficult to look at.

In Lightroom, Enfuse is the simplest approach. It works - I've used it, and it doesn't look "HDR-EY".

http://www.photographers-toolbox.com...s/lrenfuse.php

For tools in PS, others can make suggestions.

Glenn

PS: This is not an ETTR thread - it's a DR thread.

[Stagecoach]

Perhaps things are getting a wee bit off track because I believe that the ETTR question/comment originally came up when discussing determining the actual brightness of a pre-sunrise scene and how to handle this in camera, if there was a preference.

Basically, if you push the histogram to the far right are you in fact 'recording' the scene brighter than it actually is, disregarding the affect on the shadows?

Grahame

[Manfred M]

Shane ETTR is not going to do a thing for you in this type of shot, as this technique, under some very limited conditions, will potentially reduce noise in your shadows. It will not work at all in a high dynamic range image like this one.

I'm with Dan on this one; manage the highlights with a GND filter (3 stop). Because of the way you have rocks on both sides that are above the horizon line, I would be tempted to use a soft grad, rather than a hard grad, if you are doing this as a single shot. If you are looking at blending 2 (or more) shots then a hard grad right along the horizon and PP work to blend the rocks would be my preferred choice.

[Mike Buckley]

ETTR is not going to do a thing for you in this type of shot...It will not work at all in a high dynamic range image like this one.

We should keep in mind that by definition, ETTR is not possible when the dynamic range exceeds the sensor's ability to record it. That's because, by definition, ETTR never clips highlights.

[ajohnw]

Thanks for the post Shane. I'm going to remember them. I feel that you did exactly the right thing - ETTR even though it's a concept that has a number of catches which I wont pollute this thread with. As you have used it it has produced the goods.

Where you have gone wrong is in the PP. Darken and maybe ideally get some slight colour in the sky and all of the reasons for what I feel are bad crops disappear. Bad as they are trying to fix a study the image carefully rather than looking at the over all effect type problem and still don't give a decent result. One post showed the effect of darkening the sky.

By exposing to the right as you have it's obviously given good rock detail and that is the important aspect. If the sky was captured "dark" that aspect is highly likely to be more difficult to deal with. The only problem the exposure has given you is that you have an image that needs dealing with in parts - according to some. My feeling on the original shot posted was maybe balance up the light in the sky some what but that's still dealing with it parts. There are all sorts of options on a shot like this which might even involve enhancing cloud detail even adding colour to parts etc. The end results don't really have to look like some particular time of day.

The problem with ETTR as usually meant is using up raw highlight headroom as well - not easy for all sorts of reasons. It's probably best avoided as camera metering systems aren't really up to doing it reliably.

In terms of dealing with things in parts some form of selection with a blend width set is often a simple way of dealing with it or a simple painted mask that also often benefits from a short blended edge on the brush.

John
_

[Mike Buckley]

I'm really curious how someone such as Shane decides which people are correct -- those who told her ETTR is the proper thing to have done, those who told her ETTR is the improper thing to have done, and those who told her that by definition it was not possible to ETTR.

Yikes! Accordingly, I have no plans to participate in future discussions about ETTR.

[george013]

I'm really curious how someone such as Shane decides which people are correct -- those who told her ETTR is the proper thing to have done, those who told her ETTR is the improper thing to have done, and those who told her that by definition it was not possible to ETTR.

Yikes! Accordingly, I have no plans to participate in future discussions about ETTR.

That's not difficult. If there is no empty space on the right site of the histogram than there is no use for ETTR.

George

[Glenn NK]

I'm wondering if the "problems" with ETTR don't stem from a wee bit of a lack of understanding the basics of cameras and the method of ETTR?

Quoting directly from the latest dissertation by Michael Reichmann (I fixed one spelling error):

Some Caveats

Now, just to be sure that there is no misunderstanding – this approach only applies to raw files, not in-camera JPGs. Secondly, this means that you should bias your exposure towards the highlights, and the right side of the histogram. But, it definitely doesn't mean blowing the highlights.

And of course not every photographic situation will lend itself to this technique. A shot taken on a sunny day with a cloud, a mountain and a forest will challenge the dynamic range of any camera, and so there will be little opportunity of biasing the exposure toward the brighter tones without blowing out the clouds. There is no magic formula here. It's simply the physics of sensors combined with good shooting practice.

There are some excellent tools for us (hardware, software and concepts), but not all tools apply to all situations.

ETTR will not serve one well if the DR of the scene exceeds that of the sensor, and like any tool, must be understood, used correctly, and not applied to everything.

Glenn

[george013]

It's the absence of highlights that makes the use of ETTR possible. I mean highlights up to the end of the histogram.

Further I think that the article of Reichmann is not of a high quality, especial how he deals with the binary values.

George

[ajohnw]

I'm really curious how someone such as Shane decides which people are correct -- those who told her ETTR is the proper thing to have done, those who told her ETTR is the improper thing to have done, and those who told her that by definition it was not possible to ETTR.

Yikes! Accordingly, I have no plans to participate in future discussions about ETTR.

That aspect nearly caused me to not post at all. Really what exposure is all about is obtaining one which captures what is needed and also makes pp as easy and reliable as possible. In this case what is probably a jpg space metering system rather than one where some one has played with camera settings to try and include raw head room it's done it's job rather well.

John
-

[Glenn NK]

It's the absence of highlights that makes the use of ETTR possible. I mean highlights up to the end of the histogram.

Further I think that the article of Reichmann is not of a high quality, especial how he deals with the binary values.

George

George:

Specular highlights (the sun's reflections off water or a vehicle windshield) cannot be captured without some very heavy underexposure. Even the human eye/brain combination cannot discern any detail from specular highlights. Well maybe with a 10 stop ND

Non-specular highlights can be captured, and these are what one would "push" to the right end of the histogram using ETTR.


I don't understand what you mean by "not of high quality especially how he deals with the binary values".

Could you explain please? Thanks.

Glenn

So if one attempted to use ETTR on a scene with specular highlights, and tried to keep the specular highlights from blowing out, it would be a fool's errand.

[george013]

George:

Specular highlights (the sun's reflections off water or a vehicle windshield) cannot be captured without some very heavy underexposure. Even the human eye/brain combination cannot discern any detail from specular highlights. Well maybe with a 10 stop ND

Non-specular highlights can be captured, and these are what one would "push" to the right end of the histogram using ETTR.


I don't understand what you mean by "not of high quality especially how he deals with the binary values".

Could you explain please? Thanks.

Glenn

So if one attempted to use ETTR on a scene with specular highlights, and tried to keep the specular highlights from blowing out, it would be a fool's errand.

I don't mean the spectacular highlights of water but just a scene which has a white house by example and not real dark in it. This histogram wil be closed at the right site and open at the left site, and the DR of the scene is smaller than that of the sensor.

Reichmans is using an argument for his theory which I see often. I think, am sure, this is false.

A typical consumer DSLR recording 12 bits per sensel is able to record up to 4,098 separate tonal values.

    If we assume a 10 stop dynamic range this is how this data is distributed...
        The brightest stop = 2048 tonal values
        The next brightest stop = 1024 tonal values
        The next brightest stop = 512 tonal values
        The next brightest stop = 256 tonal values
        The next brightest stop = 128 tonal values
        The next brightest stop = 64 tonal values
        The darkest stop = 32 tonal values
    As can be seen, each stop from the brightest to the darkest contains half of the data of the one preceding it.

The camera doesn't record a 12 bits value or an x bit value. It's recordibng an analogue signal. Upper in this article he states that to but I think he forgot that.
In the quote above he is mixing the DR and the tonal value. That are two complete different things.
A DR of 10 doesn't give 2048 tonal values. It can be any number or binary any 2^x.
He is insinuating here that an image is the summation of individual stops. A stop is a halve or double of the light and has influence over the whole recorded value, not a part of it.

George

[NorthernFocus]

Not being an electronics expert nor digital whiz, I had previously accepted the argument about the binary values recorded for various stops of light. It is a fairly broadly used argument so I assume there is some valid basis for it. But theory notwithstaning, the main reason I accepted it is that it also compares with empirical observation. To borrow a page from Colin's book, sometimes you just have to trust your eyes. Let's consider practical observation. Take two photos of a scene with a fairly well distributed histogram, one intentionally over exposed to near clipping on the right and one intentionally under exposed to near clipping on the left. In post, when you bring the histogram back to center from the right side exposure, all of the hidden details in the highlights will be revealed. However, when you bring the histogram of the left exposed histo back to center, there will be much less detail in the darkest portions of the image. Explain it any way you like in bits and widgits, but it's easier to recover detail from highlights than from shadow. And that is before even considering noise.

While I may not implement ETTR as described/promoted by the gurus, to maximize available detail and to minimize noise, I most definately "lean to the right" with my exposure settings

[george013]

Not being an electronics expert nor digital whiz, I had previously accepted the argument about the binary values recorded for various stops of light. It is a fairly broadly used argument so I assume there is some valid basis for it.

Makes me thinking of the new clothes of the emperor.

But theory notwithstaning, the main reason I accepted it is that it also compares with empirical observation.

Underexpose for 1 stop and see if your histogram is half empty.

It's not that I say that ETTR is nonsense. Only the arguments used sound nonsense to me.

The article continues.

This helps explain why noise is seen most in the darkest areas of a file. In the brightest areas there is a lot of data and so the noise floor (which is always present) only represents a small percentage of the total signal (or data). In the darker areas, where data is sparse, ever-present noise becomes easily visible.

Isn't it easyer to say that noise is a light pixel and that light pixels are more visible in a dark background than in a light background?

George

[NorthernFocus]

...Isn't it easyer to say that noise is a light pixel and that light pixels are more visible in a dark background than in a light background?

George

Maybe easier but not accurate regarding the manifestation of noise. As stated by the other gent, there is an inherant and relatively constant noise floor for a given sensor and associated circuitry. The electronics have to separate true signal from the background noise at each photo site and then assign a value to that pixel. The higher the signal to noise ratio, the better the electronics can differentiate. In the case of our camera sensor, the signal is light. In simplest terms, pure white light is maximum signal in all three channels. Pure black(i.e. clipped shadows) is zero signal. Noise lies somewhere in between but on the lower end of the spectrum. So it is accurate to state that noise manifests itself as lower EV values where the electronics have trouble determining whether the signal received from an individual site on the sensor is actual signal or electronic noise overwhelming the signal value. So it is factual that noise statistically manifests itself more in darker regions of an image. Choosing a higher ISO setting on the camera is simply turning up the gain/sensitivity of the electronics. Higher sensitivity also raises the noise "floor" which is why higher ISO setting get noisier.

High ISO performance of modern sensors is a combination of lower noise "floors" and better algorithms to interpolate and "guess" which pixel values are noise and should be replaced with a value similar to surrounding pixels. Frankly the low light performance of current generation DSLRs is amazing compared with just a few years ago.

This is an over simplified explanation by a simple mind which really doesn't understand any of this stuff. So I hope everyone had there hip waders on prior to reading all of that

[Glenn NK]

Thanks Dan. Seems as though you've thought it through. I don't know why so many people have trouble with the LuLa article and keep quoting inaccuracies - it seems clear enough.

The following was from a post I read last week:

. . . . . because the camera loses a lot of tonal information as you get close to the highlights. One has to balance colour quality of the highlights with noise in the shadows... Generally raising the shadows produces much better results than dropping the highlights.


Glenn

[ShaneS]

Thank you for the additional comments but you guys have gone way over my head technically so I think that my best course of action is to just experiment with the camera more and when I see a scene I will try ETTR, bracketed and normal exposures for a while to see where that gets me in terms on understanding my sensor better. I will also work on trying different metering modes to see the impact of that as well. All while exploring different options in PP...

I currently do not own an ND filter of either kind and need to decide if that is a route I want to consider in the future which seems likely at some point if I continue to explore scenes like these.

Thanks to all who participated in the thread!

[Stagecoach]

Thank you for the additional comments but you guys have gone way over my head technically so I think that my best course of action is to just experiment with the camera more and when I see a scene I will try ETTR, bracketed and normal exposures for a while to see where that gets me in terms on understanding my sensor better. I will also work on trying different metering modes to see the impact of that as well. All while exploring different options in PP...

I currently do not own an ND filter of either kind and need to decide if that is a route I want to consider in the future which seems likely at some point if I continue to explore scenes like these.

Thanks to all who participated in the thread!

Hi Shane,

For info and not suggesting it is the best method I use the following procedure on these pre sunrise shoots;

a) Matrix metering selected. (although any mode will work the same)
b) Manual control with Aperture set generally for best IQ (around f/11) or higher if wanting to 'smooth' water slightly or for DOF reasons.
c) Take shot with meter centred, check histogram (possibly blinkies as well) adjust speed as necessary to put histogram where I want it.

I find the above the easiest when working with the camera low as I can see the top meter indicator easily and can make all adjustments using one thumb wheel. You very soon get used to the relationship between meter and histogram movement.

Grahame

[george013]

Maybe easier but not accurate regarding the manifestation of noise. As stated by the other gent, there is an inherant and relatively constant noise floor for a given sensor and associated circuitry. The electronics have to separate true signal from the background noise at each photo site and then assign a value to that pixel. The higher the signal to noise ratio, the better the electronics can differentiate. In the case of our camera sensor, the signal is light. In simplest terms, pure white light is maximum signal in all three channels. Pure black(i.e. clipped shadows) is zero signal. Noise lies somewhere in between but on the lower end of the spectrum. So it is accurate to state that noise manifests itself as lower EV values where the electronics have trouble determining whether the signal received from an individual site on the sensor is actual signal or electronic noise overwhelming the signal value. So it is factual that noise statistically manifests itself more in darker regions of an image. Choosing a higher ISO setting on the camera is simply turning up the gain/sensitivity of the electronics. Higher sensitivity also raises the noise "floor" which is why higher ISO setting get noisier.

High ISO performance of modern sensors is a combination of lower noise "floors" and better algorithms to interpolate and "guess" which pixel values are noise and should be replaced with a value similar to surrounding pixels. Frankly the low light performance of current generation DSLRs is amazing compared with just a few years ago.

This is an over simplified explanation by a simple mind which really doesn't understand any of this stuff. So I hope everyone had there hip waders on prior to reading all of that

I was referring to noise in general and that was wrong. He was referring to the floor noise. Excuse me. When the value of the floor noise is overruled by the signal value it disappears.
I don't say ETTR is nonsense, but the arguments he used are wrong.

George

[ShaneS]

Thanks Grahame! That is basically my set up us well and I will take shots with the histogram in different places so I can see the impact on the image.

I will likely be heading out this weekend to try again and hopefully mother nature will provide better clouds this time Lately they have been of the blobby variety which doesn't add a lot the the compositions I envision.