Plain Talk About ETTR

[BobF]

Alright, Frans asked for it, so here it is. A thread dedicated to talking about ETTR and the pros/cons of the methodology. ONLY reasoned, rational discussion permitted. Any zealotry or other unreasonable arguments presented and I'll ask the mods to shut the thread down.

ETTR - Expose to the Right - has been a controversial approach to exposing for digital photography for nigh on a decade. Perhaps the first notable writing (http://www.luminous-landscape.com/tutorials/expose-right.shtml) on the topic was by Michael Reichmann of the Luminous Landscape. He's written an update to the topic more recently that can be read here (http://www.luminous-landscape.com/tutorials/optimizing_exposure.shtml).

The theory behind ETTR goes like this: Noise is bad and should be avoided if at all possible. The majority of the signal in digital images is contained in the higher exposures and, as a result, the signal to noise ratio is improved as the exposure is increased. Ergo, in order to maximise signal and minimise noise, the approach is to push the exposure as far to the right on the histogram as possible without overexposing significant highlights. Personally, I take this further and say blowing any highlights but I recognise that certain highs, like specular highlights, are basically impossible not to blow even with other techniques such as HDR. The ETTR'd exposure will look like crap once it's brought onto the computer. What's done then is that the exposure is reduced in post to bring everything back into line with where it should be. Reducing exposure in post; however, is, in theory, better than doing it in the field because you'd still maximised your signal and minimised noise in shadow areas even after reducing exposure in post. Jeff Schewe has an article on his site (http://schewephoto.com/ETTR/index.html) that discusses ETTR. And Andew Rodney wrote an article for Digital Photo Pro (http://www.digitalphotopro.com/technique/camera-technique/exposing-for-raw.html) magazine a few years ago that, sort of indirectly, espouses ETTR. Of note, that article is still at the top of the 'Most Popular' on the DPP site.

That's the theory.

There are, obviously, limitations. ETTR, in truth, works in a few limited conditions. First, it only works if the brightness range (aka dynamic range ) of the scene fits within the dynamic range of the sensor. If that's not the case then you, as the photographer, have a choice to make. You have to choose to blow out highlights, block up shadows, or do something else like HDR or some other form of exposure blending to compensate. That's one limitation. ETTR also, for the most part, only works at base ISO. That is, it is only of value at the lowest ISO setting of the camera without invoking Lo (or however your camera denotes ISO settings below the base). Why? Simple. Let's say the base ISO of your camera is 100. You're shooting at 200 and you use ETTR to push the exposure 1 stop to the right (i.e., you've increased exposure by 1 stop from the meter). You could have achieved the same result by using ISO 100 and the same shutter speed and exposing normally. This article (http://chromasoft.blogspot.com/2009/09/why-expose-to-right-is-just-plain-wrong.html) at the Chromasoft site has a lengthy explanation. There are several other good articles related to ETTR on that site. The 'hue twist' issue is of less importance to me, personally, but it is interesting. Sandy actually has a new article in response (http://chromasoft.blogspot.com/2011/08/ettr-for-fourth-time.html) to Michael's 'Optimizing Exposure' piece.

Another reason why ETTR only works at base ISO is the drange curve of many digital cameras. The drange of the sensor is degraded as noise becomes more prevalent. Noise becomes more prevalent as the ISO is increased. Note, I didn't say there was more noise at higher ISO settings, only that the noise is more prevalent. Many of the current cameras are referred to as having 'ISOless' sensors. That is, the drange of the sensor drops in a linear fashion with an increase in ISO - a 1 stop increase in ISO leads to a 1 stop decrease in drange. The Nikon D7000 is a perfect example of such a sensor. It's not the only one. In such a case, ETTR is of no value at anything above base ISO. There are some cameras; however, where the drop in drange is less than 1 stop for a 1 stop increase in ISO. For these cameras, there may be some benefit to using ETTR at higher ISO levels. As newer cameras are introduced with better drange performance there should, actually, be more situations where ETTR could be of benefit as more scenes will fit within the range of the sensor to capture in a single exposure. ETTR also only works when shooting RAW. If you're shooting JPEG (or one of the few cameras that can capture in TIFF) or TIFF it doesn't apply because you don't have the same freedom to make adjustments post capture due to the already processed nature of the image. Once the pixels are cooked, they're cooked. Period.

Those are the few, limited, cases where ETTR may have some value. For those with a film background, ETTR actually has some connections to the film era and the Zone System. You can read my musings on that at this guest post I wrote (http://www.northlight-images.co.uk/article_pages/guest/digital_analogue.html) for Northlight Images last year.

What's noted above is the realistic truth of ETTR. It can be useful in certain, limited, situations. It is not a panacea. It is not a 'MacGillicuddy's Cure All Tincture'. It is not the elusive 'silver bullet'. It is simply one method of exposing for digital that works in some situations. Simply another tool in the box.

As stated at the outset, reasoned, rational discussion of ETTR is welcomed. Should it stray off that I'll ask the mods to shut the thread down. Clear? Good.

[JWurst]

I guess I've never really considered that ETTR only applies to base ISO, but that is how I've always treated it. It takes more light to ETTR and if I have that light available then I'd rather lower the ISO than risk blowing a highlight. At base ISO noise is not a concern of mine, so ETTR seems useless. If there was some way to deal with noise at high sensitivity levels then I'd be all for it.

[peirceman]

I found the DPP article interesting, having read it when I bought the magazine. The theory behind it isn't so much to avoid noise, but that more data is stored at the right side of the histogram as opposed to the left side of the histogram. So, continuing that thought process, if the goal of the photographer is to capture a file that captures the most data, then choosing a right leaning histogram over a left leaning histogram makes sense. Also, the article points out that using the clipping warning on the LCD makes no sense since that is a JPG conversion of your RAW data, and is not converted linearly.

Here is what makes sense to me. If I have to choose because the scene I am shooting is difficult exposure-wise, then I am going to lean the histogram as much to the right as I can without clipping the highlights. That is going to capture the most information and that helps in post. When I have used that method, it has worked for me and I like the results.

What doesn't make sense to me is to expose everything to the right, regardless of subject.

This does beg a certain question......How many of the non-ETTR zealots have actually tried ETTR and compared to to their normal way of shooting? Any thoughts?

Scott

[BobF]

It actually is about reducing the evidence of noise, Scott. That comes from maximising signal which improves the signal to noise ratio (i.e., more signal relative to the amount of noise).

There really is no other choice in the field but to use the LCD on the back of the camera. It can be used, but the photographer needs to do some testing to figure out what it's saying, how best to have the JPEG preview created and how much head room exists between the JPEG preview and the actual RAW file. It is possible to set up the camera to render a JPEG that's quite close to the RAW so that the LCD can be a reasonable approximation. Once that's done, some testing and comparing needs to be done in a RAW converter to see how much head room there is. That's pretty simple to do to.

I've used it and not used it. The differences, in most cases, are minimal and nearly impossible to discern in a reasonably sized print (i.e., up to 20x30) without really doing some serious nose-on-paper looking. I'm not sure the additional work in post is worth it. Particularly so with the new 2012 Process Version coming in LR4 and ACR which will require more adjustments to bring everything back into line. It will no longer just be an Exposure (and maybe Recovery) adjustment. It's likely to require the movement of 4 or 5 different sliders due to the manner in which Adobe have revamped the way their converters handle general exposure adjustments.

[AstroImager]

It actually is about reducing the evidence of noise, Scott. That comes from maximising signal which improves the signal to noise ratio (i.e., more signal relative to the amount of noise).



Bingo, Bob.

Scott, it's not correct that "more data is stored at the right." If you have a subtle variation in tone/color of 1 unit in an image, it takes one bit of change to register that change of 1 unit. That's the case whether the change is from a level of 10 to 11 (at the "bottom end"), or from a level of 180 to 181 (at the "top" end). The data is one bit of change, no matter where it occurs.

Say you shoot a black card, with a light gray square on it. If you expose "normally," the black area in the RAW data will average around 10, plus or minus 2-5 units because of noise (it will average around 10 because of bias current in the chip). The gray square will average around 128 units, plus or minus 2-5 units because of noise. Since the 2-5 unit variance around the 10 on the black card is a higher percentage of the average (lower signal-noise ratio), it's more noticeable to your eye than the same variance around the 128 unit average -- but they both vary due to noise exactly the same amount.
If you expose to the right, the "base" for the black goes up to, say 90, with the same +- 2-5 units, and the gray is now around 208 units, +- 2-5 units. There's no more actual data recorded in any way, there's still a difference of 118 units between the black and the gray -- but now the +- 2-5 units is a much smaller part of the "base" level for both, so the noise is less visible since it's overwhelmed by signal. No additional data was recorded in any way.
That's *all* ETTR does -- bury the noise in signal so it's less visible. Nothing else. It doesn't "record more data."

Yes, I've tried it, and do occasionally use it. I use it when I do "plate" backgrounds of skies a lot, with fluffy clouds, because they're low-contrast scenes with a DR that fits entirely in my sensor's range, and I can reduce visible noise in them by using ETTR and pulling them back in post. I've tried it for other kinds of scenes, and found no real benefit -- and in fact with the non-linearity of the top end of sensors, you lose highlight tonality far too easily. So I rarely use it.

If you expose correctly, and make conscious decisions about whether you want more shadow tonality or more highlight tonality, there's no need for it.

Paul

[WRon]

Good, sane, sensible explanation, guys. Thanks for posting this information calmly, coolly, and accurately, with no hysterics. Seems that ETTR works in some, but not all, situations, which doesn't surprise me because photography is so often that way.

[BobF]

And that's what it is, Ron; just another tool in the kit.

[nigelu]

Paul, I need to take exception to certain parts of your post.

“Scott, it's not correct that more data is stored at the right."
Scott is perfectly correct as I will show below.

“If you have a subtle variation in tone/color of 1 unit in an image, it takes one bit of change to register that change of 1 unit. That's the case whether the change is from a level of 10 to 11 (at the "bottom end"), or from a level of 180 to 181 (at the "top" end). The data is one bit of change, no matter where it occurs.”
When you change exposure, the brightness of ALL parts of the image are changed by the same percentage; if you double exposure, the brightness of ALL parts doubles; if you half exposure, the brightness of ALL parts is cut in half. Using your example, if you are at level 10, you need 1 level or a 10% change to flip one bit and get to 9 or 11; that’s a 10% change (ignoring rounding issues). If you are at level 180, you need again 1 level change to get to 179 or 181; that’s a 0.6% change. If one level changes x% as a result of exposure changes than any other level changes by the same x%. Your example doesn’t reflect the real world.

“Say you shoot a black card, with a light gray square on it. If you expose "normally," the black area in the RAW data will average around 10, plus or minus 2-5 units because of noise (it will average around 10 because of bias current in the chip). The gray square will average around 128 units, plus or minus 2-5 units because of noise. Since the 2-5 unit variance around the 10 on the black card is a higher percentage of the average (lower signal-noise ratio), it's more noticeable to your eye than the same variance around the 128 unit average -- but they both vary due to noise exactly the same amount.
If you expose to the right, the "base" for the black goes up to, say 90, with the same +- 2-5 units, and the gray is now around 208 units, +- 2-5 units. There's no more actual data recorded in any way, there's still a difference of 118 units between the black and the gray -- but now the +- 2-5 units is a much smaller part of the "base" level for both, so the noise is less visible since it's overwhelmed by signal. No additional data was recorded in any way.
That's *all* ETTR does -- bury the noise in signal so it's less visible. Nothing else. It doesn't "record more data."”
Totally agree with the noise issue. However, your numbers about before and after are completely off. I’m not going to pick a bone about your bias current number; if it is that high I would expect an in-camera correction to get a zero or near-zero value for really-really dark areas of the image. I do have a major issue with your numbers after expose to the right. The percentage increase in black value should be around the same as the percentage increase in the gray value (and even less if the bias current remains the same, which I believe to be the case). Here are my numbers as a result of real shots I just made of dark gray matte material (the darkest material I had at hand) with a medium gray center; the black is not deep-deep black, but still pretty dark.
Exposure 1: black level 38, gray level 115.
Exposure 2 at 1 stop exposure increase: black level 71, gray level 219.
As a result of increasing exposure by 1 stop, the black level, the gray level and the difference between the two all increased by a factor of about 1.9. Levels of all parts of an image change by the same percentage when exposure is changed. By increasing exposure more levels are recorded with less noise and the chance for banding or posterizing is reduced.

[Fotolopithecus]

Somewhat related to this school of thought is Nikon's metering approach on the D7000. There's a fascinating series of three short videos at testcams.com titled "does the D7000 over expose." The long, and short of it is, that it depends on whether you shoot Raws, or JPGs. He compares the two different metering philosophies of the Canon 7D, and Nikon D7000 with the results. Basically, it seems the Nikon is exposing to the right without clipping, but while thats good for Raw shooters its not ideal for JPG shooters. Anyway, you might find it interesting, and I'd be interested in what Bob, and Paul think of the guys methodology, and conclusions.;)pith

[BobF]

I'm not sure if I've seen that or not. It sounds familiar. I also seem to recall a thread at LuLa that pretty much shot his methodology all to hell. But I could well be mistaken and it could have been something else.

[Fotolopithecus]

I just went over to LULA because I'm trying to get a clear if basic understanding on the subject. Its a huge site, and finding what you're talking about Bob is proving difficult. Any idea where in there you may have read about the testcams methodology?;)pith

[BobF]

No, and after I went to look at the site you mentioned, I don't think the discussion at LuLa was about that. But.... Thom Hogan talks about it in his review of the D7000. Basically what he says is that it's not overexposure, rather that Nikon changed the metering system and there's no longer the same protection for highlights there was in older cameras. As a result, higher drange scenes will show more bias to the highlights than previously. This is actually good because it means Nikon is metering properly rather than 'dumbing down' and trying to protect the user from bad photos. Essentially Nikon has said 'this is a camera for people who know what the f^(& they're doing'.

[AstroImager]

Digital sensors don't deal in "percentages," frans. They're linear (up to the anti-bloom point). It takes *exactly* the same number of photon hits to register a 1-unit change from 10 to 11 as it does to register a one-unit change from 128 to 129. Or anywhere else in the range of values.

You also don't seem to understand the hardware; Sensors typically have a full well value of between 20,000 ADU and 50,000 ADU (ADU = analog device units). Here's how one example works:

Take a sensor with a full well depth of 50,000 ADU, with 14-bit A/D conversion as an example.
The sensor will typically have a bias current of 100 ADU (this prevents underflow errors in A/D conversion). Anti-blooming typically kicks in around 90% of full-well value (in this case, 45,000 ADU). That puts the linear range of the sensor between 100 and 45,000 ADU, a total of 44,900 possible values. 14-bit A/D conversion gives 16,385 possible digital values, mapped onto 44,900 possible analog levels, a .36 conversion rate; that means it will take roughly 3 changes in the ADU level to produce a 1-bit change in the output digital value. Basically, 3 "photon hits." That's the same no matter where it is in the range of linear values, whether the change is at the "bottom" or the "top" of the data range. And a 1-bit change is a 1-bit change, whether it's at the "bottom" or "top" of the data range. If a particular pixel during a particular exposure has already stored an ADU value of 30,000, it will take 3 more "photon hits", getting it to 30,003, to show up as a 1-bit change in the digitized output. If a particular pixel has only stored 200 ADU, during an exposure, it will still take 3 more "photon hits" getting it to 203, before a 1-bit change will show up in the output data.
At the top end where anti-blooming has kicked in, there's no longer a 1:1 correspondence between "photon hits" and ADU value stored in the pixel, because some proportion of the new "photon hits" will be drained out by the anti-blooming gates. Typically, at 95% of full-well depth, it will take around 5-8 times the usual number of "photon hits" to show a 1-ADU change, and since it takes a 3 ADU change to register in the digitized output, the highlight end is very "flat" (on a tone curve) indeed. At 99% of full well depth, usually no number of additional "photon hits" will increase the ADU value stored, they're all drained away without recording as a charge in the pixel.

Yes, adding one ADU value increase on top of 30,000 already recorded is a smaller percentage of the total -- however, it takes the exact same number of "photon hits" to make a 1-bit change anywhere in the linear range of the chip. The "percentage" is irrelevant.

Paul

[nigelu]

Paul,

Never said that sensors are not linear. As a former developer of linear amlifiers for scientific applications I have a pretty good grasp of the characteristics of linear systems. I agee with about everything in your latest post, but everyting in that post is besides the point. The point is that if you change exposure every sensor element will see the same percentage increase or decrease, not a fixed amount, of incoming photons. A simple real world test, as I described, confirms that; your exampled is fatally flawed. I suggest you pick up your camera and actually do the little test you theoritized about.

[peirceman]

Okay, maybe "data" was the wrong word to use. Here is a snippet from the article Bob referenced

In linear capture, the first stop of highlight data contains half of all the levels within the image, and the next stop, half the remaining number of levels and so on. The fewest number of levels recorded is in the last stop of shadow detail—only 64 of 4096! If the image is underexposed, even fewer levels are used to describe that last stop of tonal data. This is also where most camera noise is found, so the results of underexposure are more noise, less actual data.

So, if the authors tenets are true, then choosing to capture the highlights and sacrifice a bit of the shadow will give more "levels" of information than doing the opposite. I took "levels" to mean "data". There is still part of my brain that believes that if levels = information = data, then by logic levels = data, but I don't have enough technical knowledge to support my feelings.

With that said, the author goes on to discuss that his version of ETTR isn't over-exposing on purpose, but instead is the conscious decision to "Expose for the highlights, develop for the highlights" To me, this is not the same as setting the camera to adjust exposure automatically +1 EV or whatever.

[nigelu]

Peirceman,

You are exactly right. Levels, information, data means all the same in this regard: information retrieved from the camera sensor and available for further processing. "ETTR isn't overexposing on purpose": right on. And you are correct again that ETTR doesn't mean automatically doing anything; you evaluate the histogram, increase exposure if there is room and decrease exposure if the histogram shows blown highlights that you don't wont to be blown.

[BobF]

Which article was that, Scott?

The concept isn't to 'expose for highlights and develop for highlights'. That simply wouldn't work under ETTR theory. If you expose for highlights; that is, place the important highlight area where you want it in the histogram then you haven't, necessarily, used ETTR. ETTR is, exactly that, Expose to the Right. It is a bias toward increased exposure. Not overexposure because overexposure is a mistake.

If you simply place an important highlight detail in a particular area of the histogram then you could have other, less important, highlight areas that are exposed beyond the point of recovery. With ETTR, you place the brightest highlight that can be maintained (i.e., anything other than specular highlights) at the right edge of the histogram. Generally that will mean you'll be pulling back everything. But, depending on the scene, it could also mean you're pushing up some other highlight area and bringing down other areas. Pushing up some areas isn't good because you're losing on the S/N relationship. That's another reason the ETTR approach can be flawed and why it has to be used in moderation and with previsualisation of what you want the scene to look like after development.

It's not true that 'more' noise is found in the shadows. Noise, for the most part, is a constant. There is some variability to it, but not a lot. I'm not going to get into the technicalities of read noise and shot noise. If you want that, go over to LuLa, there are some who discuss it over there with the same passion most of us would discuss the love of our life (I am talking person). But for the most part it's a constant. Noise is 'more apparent' in shadow areas because there's less signal. The more signal (the more data, the more levels, the more photons) relative to the fixed amount of noise (because signal isn't fixed) means that noise is less apparent in the highlight areas but there's no less of it in nominal terms. It's a relative equation between signal and noise.

[peirceman]

Which article was that, Scott?

That is a direct cut and paste from the article you posted the hyper-link for in Digital Photo Pro Magazine.

Scott

[JWurst]

Nigelu, Paul,
Okay, so exposure is a percent sort of thing--doubling the size of my aperture doubles (200%) the light on all areas, which is a much larger increase in photons in the highlights than in the shadows. But then it all gets stored on a linear sensor and saved into a linear file (128 is middle in an 8 bit file). So on what actually hits the sensor and becomes the highlights there must be a lot more...I don't know what to call it--resolution of light intensity. In past ETTR arguments I've heard plenty about how much more data there is there. But then when it gets stored most of that must get discarded when it all gets smashed into the top 128 possible values of the high key half of the pixel value range. On the low end a much smaller "resolution of light intensity" would get smashed into the same amount of space in the low key half of the pixel value range. It seems to me that the only way ETTR would be beneficial, other than for noise, would be if the "resolution of light intensity" at the low end were actually smaller than the output file could resolve, resulting in gaps in the histogram. I'm pretty sure that isn't happening when I convert the RAW to an 8 bit file. I'm not sure about 16 bit. I tried to test and got identical results in each. I don't think I really know how to test.

[BobF]

That is a direct cut and paste from the article you posted the hyper-link for in Digital Photo Pro Magazine.

Scott

OK, thanks. That makes sense. He's dumbing it down a bit for the audience.

JW, there is no other benefit to ETTR than to try an maximise S/N and reduce the appearance of noise. That's it. That's all ETTR is supposed to do. None of the data gets thrown away as long as you don't go into overexposure. In editing, none of that data gets thrown away either. It simply gets remapped to other areas.

[peirceman]

OK, thanks. That makes sense. He's dumbing it down a bit for the audience.

LOL, thanks Bob. I guess with my limited knowledge I am part of the "audience".

[ChanTran]

WOW! so all these years of shooting color negative film I was ETTL without knowing it. I ensure my shadow areas has sufficient exposure to have good details and not grainny mud but not more. I generally let the highlight goes where it may and fix it in the post. Oh no actually printing them highlight down so they are at correct density on the print.

[WRon]

Chan:

Yep, we negative film shooters exposed for the shadows and developed for the high values. In short, we gave the film enough exposure to ensure that it captured important shadow details and then developed the film just long enough to ensure that the important high values didn't block up. I'm pretty sure that ETTR is not the same thing.

Ron

[BobF]

You did that if you shot b&w, Ron. Not if you shot colour neg film.

Read the link I posted of my guest post on Northlight Images. More similarities than you might think. ;-)

[WRon]

You did that if you shot b&w, Ron. Not if you shot colour neg film.

Read the link I posted of my guest post on Northlight Images. More similarities than you might think. ;-)

Right on, Bob. Color neg was a horse of a different color. I'll check out the post on Northlight.

[nigelu]

Bob, thanks for your original post, which was a good primer on ETTR with a wealth of references to articles by others on the subject, with one caveat: you didn't include the issue that many people claim that increasing exposure increases the number of captured levels and decreases the likelihood of posterization in post-processing and that others, in the minority, claim that it doesn't. It would have been even better if you would have included your opinion on that issue.

Paul, in the spirit of Bob's request for a reasoned, rational discussion, it would have been nice, to say the least, if you had followed up on the issue whether or not increasing exposure increases the number of captured levels and either proved me wrong or admitted that what I said was correct. So far, you didn't prove me wrong; all you did was describe the linear response of solid-state sensors, something we agree on.

[BobF]

As often happens with any kind of theory, people will try to connect other issues to it and say 'well, see if you do ABC, you won't get, XYX.'

There are a few main ways you can get posterization in an image. One is from editing low bit depth images. There just isn't enough information in the image to 'fill in' when you start pushing pixels around. Another is if you try and push an area too far - particularly shadow areas and trying to push them higher.

So sure, ETTR may be helpful in reducing posterization from trying to push up shadows. But the number of times that's going to be of value are small. Further, if you're trying to push up low values so much that you're causing posterization then the image probably wasn't exposed properly in the first place. It was probably underexposed. I can count on one hand and have fingers left over the number of times I've had trouble with shadow posterization from trying to push up those areas. Expose properly and it won't likely be an issue. With RAW files, you do a fair bit of manipulation without bringing in posterization just due to the nature of what RAW files are.

If you're editing low bit depth images then it was probably shot as a JPEG and ETTR doesn't apply anyway in that case. So here again, ETTR really isn't necessarily the answer. Editing high bit depth images is more likely the answer. Or at least an alternative answer in many cases.

You can also introduce posterization via ETTR if you happen to clip a channel. This article on LuLa (http://www.luminous-landscape.com/tutorials/restore-clipped.shtml) discusses it.

So to make a blanket suggestion that ETTR will solve issues of posterization is far too broad a statement. It could but it doesn't necessarily or automatically and the reasons for posterization being a problem quite possibly point to a different solution.

[nigelu]

Bob, you are right that editing low-bit images can cause posterization. For high-bit images I don't think that pushing shadow areas too far is the most common cause for posterization. The far greater threat is moderate image editing which can result in posterization in smooth areas with only moderate tonality and color changes like blue sky; that's where posterization will show up first, even when doing only a moderate amount of editing like pulling up the whitest white areas in your image, something that you would avoid if using ETTR correctly. Starting out with more levels through the use of ETTR will reduce the chance for posterization to show up in those delicate areas.

Furthermore, I didn't make a blanket suggestion that ETTR will solve posterization issues; I said that it would reduce the chance of posterization.


Paul, feel free to come back any time and prove me wrong on the issue of whether or not ETTR increases the number of captured levels!

[stargate87]

Nice one Bob. I love a good ETTR thread. :cool: Its like an oil thread on an automotive website.

[BobF]

We'll disagree that ETTR is a solution for that type of situation. I didn't say pushing shadow areas was the "main cause". I said trying to push "an area too far - particularly shadows". That means pushing pixels in other areas can lead to posterization as well. In point of fact, noise can be effective in hiding banding/posterization and increasing the S/N may contribute to the problem. While Guillermo and I don't agree on a number of things, his explanation here (http://www.luminous-landscape.com/forum/index.php?topic=59613.0) may be of interest.

Randy - Synthetic. :-)