rmalarz Loc: Tempe, Arizona

Yes, I'm well aware of all of this. The reason I asked the question was that your writing only referenced specifically changing shutter speed. Since exposure is solely dependent on shutter speed and aperture, the isolation of shutter speed inspired my question.
--Bob

big-guy Loc: Peterborough Ontario Canada

<sigh> so by your own admission, my original statement stands, they are "basically" the same thing just coming at it from different aspects. ETTR brings you to the door/threshold without going beyond the right and DBTH is a guard at the door/threshold making sure you don't go beyond the right.

selmslie Loc: Fernandina Beach, FL, USA

The difference is that with DBTH the histogram does not need to get close to the right. It can be right in the middle.

If the scene's DR is low and the JPEG straight from the camera looks right, the histogram may not be close to either the right or left end of the range.

But in that case the ETTR version will be touching the right end. The JPEG will look washed out and you will need to reduce the Exposure (gain) in your raw editor. That's extra work.

selmslie Loc: Fernandina Beach, FL, USA

Here is a tabular comparison between a normal exposure and the EBTR version for scenes with three different dynamic ranges.

The EBTR approach aligns the brightest stop with the greatest number of possible encoded values regardless of ISO. For a 14-bit raw file that is 8192.

The normal approach aligns middle gray with the "middle" of the range, 512 values. Both of these values are far greater than the uniform spread of 8-bit values 256/9, about 28.5 values per Ev. For 16-bit values it is about 1820 values per Ev.

The raw conversion essentially transforms the arithmetic progression of raw values into a logarithmic progression, which is how we normally see a progression of tones from black to white. After the conversion the number of values is pretty much evenly spread across the DR - until you continue to edit the image.

For a scene with a DR of 9 Ev, both EBTR and normal exposure are pretty much the same. The difference in the distribution of values only shows up when the DR is less than 9 Ev.

As the attached spreadsheet shows, EBTR is just a numbers game. During the raw conversion, the values in the EBTR image are converted logarithmically the same way as for the normal exposure except that the added exposure has to be backed out.

Both approaches exceed the threshold for a human to be able distinguish between tonalities throughout the 8-bit range. The only difference is that, when converting to a 16-bit image, where the number of potential values drops below 1820 per Ev, the EBTR version begins to do something different one or two stops later than the normal version.

Sure, the numbers are different. But the point is that the difference is impossible to see or demonstrate in a real image. That's why no proponent of ETTR/EBTR is willing or able to post any real images to show the visible difference along with a procedure for others to replicate the process.

(Download)

selmslie Loc: Fernandina Beach, FL, USA

Would that be the pot calling the kettle black?

No, my mind is not closed. I went through the entire exercise, tried ETTR, EBTR, measured my ERADR at various ISO settings. What did it get me? Absolutely no visible benefit in any image.

I then did a lot of research and experimentation to find out why. I have published the results of those experiments giving documentation, procedures and a host of examples.

All of my evidence is presented honestly at full resolution so anyone can examine it closely. I have only provided screen shots and tabulated data in support of this evidence.

What have you presented in therms of evidence? Absolutely nothing! No images, only crops and screen shots of images taken using unreasonably high ISO or very low light levels.

You have spent an inordinate amount of time discussing ERADR and how to measure it but is it really necessary to know it within 1/3 of a stop? Probably not. It's likely somewhere between one and two stops but what difference does that really make in the real world.

The brightest Ev step in you image is going to occupy a range of less than 30 units in an 8-bit image. Does it really matter whether the raw file starts off with 8000, 4000, or even 500 distinct values that are going to get boiled down to 30 units in the JPEG? Of course not. So EBTR provides absolutely no visible benefit in the highlights.

If there is going to be any visible benefit at all it will need to occur at the other end of the scale, in the shadows. But it has been impossible for you to demonstrate that with a real image at base ISO!

So which one of us is being open minded and honest?

Go ahead. Accept my challenge, honestly and with an open mind. At base ISO. Do you have the courage of your convictions?

altheman Loc: Christchurch, New Zealand

I have used ETTR etc for quite some time and because I have a reasonably standardised workflow it was no problem to create presets in Lr to compensate for the "overexposure" and I know that there is a definite reduction in noise when using this workflow.
What I have noticed with modern sensors is the reduced need for HDR bracketing and although when the dynamic range of a scene is challenging I may still bracket for HDR there are many times when I don't blend the images because there is no need.

selmslie Loc: Fernandina Beach, FL, USA

I don't know if you have followed my entire presentation or if you just started here so I will reiterate a couple of points.

The visibility of noise in an image is reduced only be increasing the exposure (opening the aperture and/or slowing the shutter). That increases the signal and, since base noise is for all practical purposes a constant, it increases the signal to noise ratio (S/N) and reduces the visibility of noise.

For a given exposure, ISO has nothing to do with S/N or the visibility of noise and I have demonstrated that here and several more times here. The reason changing ISO does not affect S/N is that both the signal and base noise are amplified by the same amount.

Whether you get to your final exposure using ETTR/EBTR or conventionally, if you end up with the same exposure you will end up with the same S/N. The only difference will be the location of the histogram.

Here is a simple example you can try to prove it to yourself:

Image #1 - Set your ISO to 800 and meter the scene. Suppose the camera recommends 1/1000 @ f/11. You find that the histogram will end up three stops to the left of where you want it so you add three stops of exposure by setting the shutter to 1/125. You could open the aperture instead but let's assume you don;'t want to mess up your DOF.

Image #2 - Set your ISO to 100 and meter the scene. The camera will recommend 1/125 @ f/11, the same as your EBTR exposure in #1. The histogram will be in the same place it was initially in the ISO 800 version.

Image #3 - Using the same exposure of 1/125 @ f/11, change your ISO back to 800. The histogram will now end up in the same place it ended up in Image #1.

Images #1 and 3 will have the same appearance on your LCD and their histograms will be the same. All three images will have the same amount of visible noise and, after you back out the +3 Ev for #1 and #3 in your editor, they should all look identical - if you didn't blow the highlights in #1 and #3.

So based on the evidence I have presented, ETTR/EBTR does not reduce noise. Only increasing exposure reduces noise. You can get that same reduction in noise by reducing the ISO (until you end up at base ISO) and exposing normally.

selmslie Loc: Fernandina Beach, FL, USA

I cannot respond on Uuglypher's thread because he has blocked me (he is not blocked here).

My series of threads is simply a presentation of evidence regarding the efficacy of rigorous adherence to the practice of ETTR/EBTR. The exercise of ETTR/EBTR clearly requires more effort than simply using the lowest practical ISO, following the camera's recommended exposure and being aware of the highlights.

I recently posted an example of shadow recovery that shows how much you can actually get back. An even more drastic example is Don't Give Up on Dark Images. In both examples I used ISO 200 and had no issues with noise.

So if we rule out noise reduction as a reason to use ETTR/EBTR, what is left? This challenge is about getting someone to come up with another benefit. So far nobody has come forward.