Debunking ETTR

[joe_c5]

<blockquote>

<p>.....and still no mention of "DTTL".<br /> ....or an explanation of the dreaded color shift.</p>

</blockquote>

<p>Without Developing to the Left, most of the images exposed to the right would just appear overexposed, so it did not seem necessary to mention.</p>

<p>Color shifts are caused by nonlinear transformations to the image <em>before </em> the exposure is corrected. If the exposure is corrected immediately on the linear raw data, the colors will be the same as in a normal exposure no matter what post processing is done after the exposure correction. In the case of that rainbow disappearing, it had just been clipped in the camera (and clipping is nonlinear).</p>

<p> </p>

[william l. palminteri]

<blockquote>

<p ><a href="../photodb/user?user_id=4977907"><em>Joe C</em></a><em> </em><a href="../member-status-icons"></a><em>, Mar 16, 2010; 03:41 p.m.</em></p>

 

<p><em></em></p>

<blockquote>

<p><em>.....and still no mention of "DTTL".<br />....or an explanation of the dreaded color shift.</em></p>

</blockquote>

 

<p><em>Without Developing to the Left, most of the images exposed to the right would just appear overexposed, so it did not seem necessary to mention.</em></p>

</blockquote>

<p>Joe, you're right of course, but this thread is the first time I've heard the topic talked about without the corollary.<br>

It's always been ETTR - DTTL.</p>

<blockquote>

<p><em>Color shifts are caused by nonlinear transformations to the image before the exposure is corrected. If the exposure is corrected immediately on the linear raw data, the colors will be the same as in a normal exposure no matter what post processing is done after the exposure correction. In the case of that rainbow disappearing, it had just been clipped in the camera (and clipping is nonlinear).</em></p>

</blockquote>

<p>This deserves a topic unto itself. Thanks for stepping up, Joe.</p>

<p>Bill P. vs. the Forces of Doom</p>

[ben_goren]

<p>Joe C wrote:</p>

 

<blockquote><p>Color shifts are caused by nonlinear transformations to the image before the exposure

is corrected.</p></blockquote>

 

<p>And therein is revealed a major practical shortcoming of ETTR. At least ACR and LR, and I think

(but don’t care to check) DPP don’t allow one to apply ETTR-specific corrections before

proceeding with the rest of the workflow. Basically, you need to do some sort of linear RAW conversion

and do all your other work in straight Photoshop, or something else like that.</p>

 

<p>And, frankly, as far as I’m concerned, being able to use ACR is <em>far</em> more

beneficial to image quality than one less stop of noise (and, of course, one less stop of headroom).</p>

 

<p>Cheers,</p>

 

<p>b&</p>

[dan_south]

The in-camera histogram is not reliable in determining clipped highlights.

 

Brilliant! This makes ETTR seem like some sort of absurd irony.

 

The R in ETTR stands for 'RIGHT', so named to signify the RIGHT SIDE of the HISTOGRAM. But when you read the fine print you find out that your sRGB-based histogram is completely inept at representing the limits of sensor latitude. So, it's not REALLY ETTR. It's more like ETTR with a fudge factor (ETTRWAFF), a fudge factor that's difficult to calculate and impossible to verify in the field (unless you're shooting tethered to a laptop and you have a lot of time to analyze data between exposures).

[jeff_schewe]

<blockquote>

<p>Dan South wrote:<br>

The R in ETTR stands for 'RIGHT', so named to signify the RIGHT SIDE of the HISTOGRAM. But when you read the fine print you find out that your sRGB-based histogram is completely inept at representing the limits of sensor latitude. So, it's not REALLY ETTR. It's more like ETTR with a fudge factor (ETTRWAFF)</p>

</blockquote>

<p>Yep...if your scene has highlights at risk of clipping...or you can use a spot meter to determine the scene contrast range and test your own sensor to determine the dynamic range and approach exposure intelligently-if the contrast scene is over a stop lower than the dynamic range of your sensor, you can ETTR a stop with little risk....or, shoot with an auto-bracket if there a question. No need to over dramatize the exercise...</p>

<p>Actually, doing an auto-bracket can make a huge impact on increasing the shadow detail while still keeping all your highlight detail by either doing HDR or process the out the exposures as layers and do layer blending to apply the lighter exposure in the shadows and the darker exposure in the highlights–which is what I tend to prefer other than the full blown HDR approach which tends to produce unnatural tone curves...</p>

[joe_c5]

<blockquote>

<p>And, frankly, as far as I’m concerned, being able to use ACR is <em>far</em> more beneficial to image quality than one less stop of noise (and, of course, one less stop of headroom).</p>

</blockquote>

<p>I don't think I would recommend using expose to the right for an entire workflow, no.</p>

<p>However, in theory software could be designed to support it. I need to figure out a way to pitch things to Adobe.</p>

<p>Going even further, cameras, especially with Live View, could calculate the exposure automatically and write the amount of correction needed into the metadata. That would take more development time than changing the order of operations on Lightroom's sliders, but it is possible.</p>

<p> </p>

[joe_c5]

<blockquote>

<p>or process the out the exposures as layers and do layer blending to apply the lighter exposure in the shadows and the darker exposure in the highlights–which is what I tend to prefer other than the full blown HDR approach which tends to produce unnatural tone curves...</p>

</blockquote>

<p>HDR, by itself, has nothing to do with unnatural tone curves. The problem is that most people (and most software?) seem to insist on applying unnatural tone curves to their HDR images.</p>

[jeff_schewe]

<blockquote>

<p>Joe wrote:<br>

The problem is that most people (and most software?) seem to insist on applying unnatural tone curves to their HDR images.</p>

</blockquote>

<p>I'm perfectly willing to assign a lot of blame to the software because most of that stuff is designed for combining way too many images and then not making it easy for the user to apply a gentle more subtle tone curve...</p>

<p>For me, if I've done a 3 shot 1.33 or 1.66 bracket, processing the lighter image for the shadows and the darker image for the highlights and blending then into the normal exposure can be done using layer blending options and layer masks to localize the tone adjustment just to areas that need them.</p>

[ben_goren]

<p>So, to recap:</p>

 

<ul>

<li>ETTR, even in theory, can give no better results than shooting at an equivalent lower ISO.</li>

<li>Every stop of ETTR used also decreases the highlight dynamic range by one stop.</li>

<li>No on-camera tool is capable of accurately indicating when highlights are clipped.</li>

<li>Even if camera firmware were to be modified to show histograms based on RAW data, there still

would be practical problems in indicating the clipping of small, subtle highlights such as those in Joe

C’s picture of the glass above.</li>

<li>No popular RAW conversion software is capable of applying ETTR in a typical creative workflow

without introducing color shifts and other undesirable artifacts.</li>

<li>Noise reduction and shadow recovery can significantly mitigate many of the image quality problems ETTR is

designed to address.</li>

<li>The tonal ranges where ETTR’s detail-enhancing effects are most visible are the same

ones very commonly crushed to Zones I and II.</li>

<li>Nobody except ETTR evangelists seem to be upset with the image quality of modern DSLRs

at base ISO.</li>

</ul>

 

<p>That about sum it up? Because, if so, we’re left with a theoretical marginal improvement in

image quality beyond outstanding that’s damned impractical to reliably implement in the real

world. Which, as I recall, is exactly what I stated at the beginning of this thread.</p>

 

<p>Cheers,</p>

 

<p>b&</p>

[joe_c5]

<blockquote>

<ul>

<li>ETTR, even in theory, can give no better results than shooting at an equivalent lower ISO.</li>

</ul>

</blockquote>

<p>I have to dispute that first point. At the lower ISO settings, DSLR cameras tend to be limited more by read noise than shot noise, so a photo at ISO 100 will not have an entire stop more dynamic range than a photo at ISO 200, probably not even close to a stop.</p>

<p>At higher ISOs, say 1600, that statement would become more or less correct.</p>

<p>Also, points 3 through 5 are true, but could in theory be fixed, or at the very least improved.</p>

<p>The last point is interesting. I can see a one stop difference at 100% zoom on my LCD monitor, but that may be because its black level is so poor. Really the shadows are supposed to be darker.</p>

[digitaldog]

<blockquote>

<p>I have to dispute that first point.</p>

</blockquote>

<p>As would I and many others.</p>

<blockquote>

<p>Also, points 3 through 5 are true, but could in theory be fixed, or at the very least improved.</p>

</blockquote>

<p>Agreed although the further sloppy reporting of the OP fails to mention that point 3 isn’t correct in terms of JPEG and is in terms of raw. Shooting for JPEG and expecting the raw data to behave the same is why we need this ETTR term and technique. Also, in terms of point 4, one could, if supplied the linear histogram, avoid all actual raw clipping so its kind of moot. <br>

All points past 5 are equally bogus, no reason to go there, it would only guarantee another 50 posts from the OP. I can’t blame you for not pointing them out ;-) </p>

[jeff_schewe]

<blockquote>

<p><a href="../photodb/user?user_id=820080">Ben Goren</a><br>

So, to recap:<br>

:::snip:::<br>

That about sum it up?</p>

 

</blockquote>

<p>Uh no...I guess you really haven't learned a thing (which considering the amount of time you've spent is a shame)...</p>

<p>To sum things up, you STILL don't understand ETTR...you don't understand digital sensors, you don't know how to use a meter to measure scene contrast range (and have no idea what the dynamic range of your sensor is) you really don't understand how to use either Camera Raw or Lightroom raw processing to control tone and color, you are mildly correct regarding 3rd party noise reduction (I prefer Noiseware but ACR 6 and LR 3 are bound to advance the noise reduction) and you are still full of crap regarding "ETTR evangelists" because #1, you still don't understand ETTR...</p>

<p>That pretty much sums it up...and the sad fact you seem your have your head in the sand regarding LEARNING...you really don't seem to want to advance your knowledge...well, ok then.</p>

<p>So, to reiterate...if you are concerned with image quality, and are shooting a scene whose contrast ranges is lower than your sensor, you can expect a noticeable improvement in image quality if you deploy the "Expose To The Right" approach to increasing (not over exposing) your digital capture exposure (read: allow more photons to fall on the sensor). This has been proven repeatedly (even by the OP's own examples) throughout this thread...</p>

<p>Since I kinda care about image quality, I'm perfectly willing (and happy to do so) to take advantage of any and all technique that improves my image quality without costing a ton of time or money...heck, I'm even willing to drop big bucks to move up from DSLRs to a P-65+ Phase One medium format back in order to improve my image quality (and ya know, ETTR works with MF backs just as well as DSLRs).</p>

<p>I'm used to making really big prints and examining my prints closely–my standard print size is 24"x30" on Epson Exhibition Paper from my Epson 9900 printer...and if you take care with the shooting (a tripod is your friend) and exposing and processing, you would be amazed at the quality you can get out of DSLRs (let along a 60 MP back...)</p>

<p>In fact, you can see some of the shots I've put together at: <a href="http://schewephoto.com/SW-Selects/"><em>SW-Selects</em></a> many of which are multiple image pano assemblies from a shoot in Southern Utah...most of which were done with multiple images shot on a bracket with the higher exposure blended into a lower exposure...with the exception of the Antelope Canyon shots which alas, my multi-shot bracketing screwed up and I ended up doing multiple shots of the same darn exposure...so, fortunately, the dynamic range of the camera back (about 12 stops) and my metering capability allowed me to pull a lot of shadow detail out without a ton of noise...</p>

 

[mirek_elsner1]

<blockquote>

<p>Andrew Rodney wrote:</p>

<blockquote>

<p>Which is still based on exposing for film with an H&D curve, not a linear capture.</p>

</blockquote>

<p>Gee, I guess those engineer types really must be damned stupid to not have figured out yet that they’re making digital cameras.</p>

</blockquote>

<p>Most likely they aren't stupid, but implementing ETTR is not that easy, I think. Correctly exposing to the right means that you have to evaluate all highlights and make sure you don't clip them. This is easy to evaluate after exposure by looking at the histogram or the blinkies, but to get it right before exposure this would probably require more than a few metering segments. And if you manage to expose it properly (i.e. as much as possible to the right, but so that you don't clip anything), the next problem is that the in-camera jpeg processors would have to have different algorithms to produce results with correct lightness.</p>

[peter_stacey1]

<p>Dan South wrote:</p>

<blockquote>

<p>I guess that would depend on the location of the highlights in the frame. If the highlights are in the center or over the currently activated focus point, they'll have more of an effect on the meter than if they fall on the edges of the frame.<br>

My D700 regularly burns out highlights in Aperture Priority Mode/Matrix Pattern with no exposure compensation applied. I'm not talking about clipping the histogram; I'm talking about areas of unrecoverable detail in the RAW file.</p>

</blockquote>

<p>Sorry to take a day to get back to you.</p>

<p>In relation to the position of the highlights, it doesn't really matter in my experience (D3, D2, D200). Matrix metering adds additional weight to the centre, so if there are more highlights in the middle of the frame, I would normally expect the camera to slightly underexpose; and more likely to slightly overexpose if the centre is filled with dark tones. But matric metering is good when I'm not using an incident light meter (as is in camera spot metering) and todays cameras have a lot of information on-board to help.</p>

<p>In practice, I find that adding +2/3 exp. comp when I'm not manually metering a scene doesn't clip the data.</p>

<p>I manually meter ~80% of the time, so perhaps one day a week I'm in a situation where I'm matrix metering rather than using an light meter (both for incident metering and spot metering).</p>

<p>But that's my experience and photography isn't a team sport, it's an individual pursuit. I'm sorry to hear about your experience with the D700, but it seems you know your system well and your the only person in a good position to know how it responds to different situations.</p>

<p>I wonder whether in certain situations, bracketing might be a good approach with your system. In film days, bracketing was a standard practice in certain situations, so that after processing you could pick the best exposure for printing (particularly with slide film).</p>

<p>Coming back to the topic of the thread, ETTR is just the same thing, but instead of using the autobracketing feature of the camera, it's an approach that involves manual bracketing of shots because we can have instant feedback on the data (although imperfect', still useful).</p>

<p>You don't seem to particularly like ETTR, and that's OK, but perhaps autobracketing in camera is something you are more comfortable with, and it's really just another form of the same thing.</p>

<p>Regards,</p>

<p>Peter</p>

[digitaldog]

<blockquote>

<p>Correctly exposing to the right means that you have to <strong>evaluate all highlights and make sure you don't clip them. </strong></p>

</blockquote>

<p>Something many of us did with transparency film for years and years (and did so for a living to boot). </p>

<blockquote>

<p>This is easy to evaluate after exposure by looking at the histogram or the blinkies, but to get it right before exposure this would probably require more than a few metering segments.</p>

</blockquote>

<p>We didn’t have histograms or blinkies in those days either. Surprisingly, we managed to expose correctly for the media. It wasn’t, and still isn’t rocket science. </p>

<p>Shockingly enough, some of us actually captured images without autofocus, using sheet film, working swings and tilts. In photo school, the first half year we shot <strong>nothing</strong> but B&W 4x5, and were not allowed to burn or dodge any of our prints (w/proof sheet on the back of the board to prove it), so we learned to expose and light properly. The idea of a histogram and on camera LCD was about as expected as traveling faster than the speed of light! </p>

[digitaldog]

<p>Will, you’re a classic:</p>

<blockquote>

 

<blockquote>

<p><em><strong>Since you prefer noise, then yes, don’t expose optimally. In fact, under expose.</strong></em></p>

</blockquote>

<p>....is behavior unbecoming of his supposed "expert" status in the industry and is antagonistic,<strong> NOT the other way around..</strong></p>

 

</blockquote>

<p>Lets look at the facts in the post and see who is acting like a potty mouth child, your new buddy the OP when he wrote:</p>

<blockquote>

<p>Tell me, Andrew, have you stopped beating your favorite underaged male prostitute? <br>

That’s probably the best characterization of a measurebator I’ve ever come across.</p>

</blockquote>

<p>How is that helpful?<br />And you think that we're fortunate to have children like that here on Photo.net?<br />Okay then, have it your way Will....</p>

[peter_stacey1]

<p>Andrew Rodney wrote:</p>

<blockquote>

<p>Something many of us did with transparency film for years and years (and did so for a living to boot).</p>

</blockquote>

<p>And with only 4-5 stops of lattitude as well. Much less than modern digital SLRs at low ISO values, yet somehow we all got by.</p>

<p>I'm glad that I've shifted almost exclusively to digital (sometimes I still have to shoot with my old F3 because it can operate without power, which is a restriction I occasionally face) because techniques like ETTR couldn't be applied back then. Having them available gives me more options to produce good images, not less, which seems to be the basic idea behind the start of the thread.</p>

<p> </p>

[william l. palminteri]

<blockquote>

<p ><a href="../photodb/user?user_id=361342">Andrew Rodney</a> <a href="../member-status-icons"><img title="Subscriber" src="http://static.photo.net/v3graphics/member-status-icons/sub9.gif" alt="" /><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></a>, Mar 17, 2010; 10:22 a.m.</p>

 

<p>Will, you’re a classic:</p>

 

</blockquote>

<p><em><strong>You bet !</strong></em></p>

<blockquote>

<blockquote>

<blockquote>

<p>

<p><strong></strong></p>

<em>Lets look at the facts in the post and see who is acting like a potty mouth child, your new buddy the OP when he wrote:</em></p>

</blockquote>

</blockquote>

<blockquote>

<p><em>Tell me, Andrew, have you stopped beating your favorite underaged male prostitute? <br />That’s probably the best characterization of a measurebator I’ve ever come across.</em></p>

</blockquote>

<p><em>How is that helpful?</em></p>

</blockquote>

<p>Where did I say it was helpful?</p>

<blockquote>

<p><br /><em>And you think that we're fortunate to have children like that here on Photo.net?</em></p>

</blockquote>

<p>Where did I say that?</p>

<blockquote>

<p><br /><em>Okay then, have it your way Will....</em></p>

</blockquote>

<p>What way is that, Andy?</p>

<p>Bill P. vs. the Forces of Doom</p>

[digitaldog]

<blockquote>

<p>What way is that, Andy?</p>

</blockquote>

<p>First, attributing quotes to me I didn’t make, then calling the kettle black. Not that I really care, just thought I’d point out the hypocrisies.</p>

[ben_goren]

<p>Ben Goren wrote:</p>

 

<blockquote>

 

<p>Andrew Rodney wrote:</p>

 

<blockquote>

 

<p>Which capture has better data from the same equipment, the one where you under exposed or

the one where you properly exposed for the raw data? A simple yes or no answer is fine.</p>

 

</blockquote>

 

<p>Tell me, Andrew, have you stopped beating your favorite underaged male prostitute? A simple

yes or no answer is fine.</p>

 

<p>Andrew Rodney also wrote:</p>

 

<blockquote>

 

<p>Practicalities have nothing to do with the science of ETTR.</p>

 

</blockquote>

 

<p>That’s probably the best characterization of a measurebator I’ve ever come across.</p>

 

</blockquote>

 

<p>Andrew, there’s a word for somebody who berates people with loaded questions and then

cries like a stuck pig when called on it. It’s the same word for somebody who preaches the

vital importance of the application of theoretical technical perfection to the detriment of artistic expression and then expresses

righteous indignation when the common slang term for that philosophy is applied to him.</p>

 

<p>That word is: hyper…um…hepta…no….</p>

 

<p>Ah, yes. That’s it. The word is, “hippocondriac.” Yes, I’m quite certain

that that’s the word.</p>

 

<p>Of course, there’s another word entirely for somebody who quotes people out of context to

score debating points, but my remembery seems to be failing me and I can’t for the life of me

think of what that word might be.</p>

 

<p>Cheers,</p>

 

<p>b&</p>

[peter_stacey1]

<p>Surely Ben, we can just get this back on track and forget the personal stuff.</p>

<p>Ideally, the morally disguting comment made earlier in the thread should be withdrawn and an apology offered. It truly was awful and doesn't belong on the forum.</p>

<p>If the personal stuff is taken out, then perhaps it can get back to a discussion of the merits of the technique and not a discussion of the merits of the individuals.</p>

<p>There is always something to learn in any thread, even ones as tiring as this one and it would be a pity for it to be closed on the basis of personal attacks.</p>

<p> </p>

[mike_hitchen]

<p>Ben - can I just give a real-world example of how I use ETTR and can you say if I am being misled on its advantages.<br>

I have often (as I am sure many have) shot a landscape scene and the histogram has shown a range of exposure from relatively dark highlights on the left spreading to about 2/3 across to the right. A fairly low-key image. I decide that I will overexpose by about a stop to bring the exposure towards the right hand edge of the histogram. In fact, experience tells me that I can then add another 2/3 to one full stop (beyond the right hand edge of the histrogram) <em>without blowing any highlights</em> because the histrogram is showing me the JPEG data, not RAW data. So I overexpose by up to 1.2/3 stop for one reason only - to reduce the S/N ration in the shadow areas to reduce noise. But given that I do not want to blow highlights (or, more exactly highlight detail that I want to keep) I only bother doing this where there is sufficient gap in the histogram between the end of the curve and the right hand end of the histogram: anything less than one stop I would not bother with.</p>

<p>Am I right to understand that you agree it will have an effect, but the benefit (with regards detail in the shadow areas) is so negligible that it is not worth even messing with. Do you mean it is not worth doing <em>at all</em> or that if your intention is to print then it will hardly be noticed?</p>

<p>I understand why you say it only applies at base ISO for the reasons you stated, but have not botherred to try that out yet.</p>

[ben_goren]

<p>Mike,</p>

 

<p>I think Joe C’s picture waaaaay up above of the water glass on the oak table is an excellent

example of why I consider ETTR to be disadvantageous even in situations such as the ones you

describe.</p>

 

<p>ETTR eliminates the possibility of serendipitous discoveries of miniscule highlight detail such as

that. And, so often, it’s exactly those sorts of details I find most compelling. Even

“bland” subjects such as sandstone are filled with such details. In my shot of the i1 case

above, you can see that the specular highlights of the foam are not pure white, but indeed quite

colorful.</p>

 

<p>If you’re certain that you would prefer less shadow noise to such highlight detail, ETTR

would be a good choice. I’m just having a hard time understanding what’s so

objectionable about the noise of modern DSLRs at base ISO that so much effort is put into reducing it

even further, never mind what other consequences it might have.</p>

 

<p>Cheers</p>

 

<p>b&</p>

[william l. palminteri]

<blockquote>

<p ><a href="/photodb/user?user_id=4564852"><em>Mike Hitchen</em></a><em> </em><a href="/member-status-icons"><em><img title="Subscriber" src="http://static.photo.net/v3graphics/member-status-icons/sub2.gif" alt="" /></em></a><em>, Mar 17, 2010; 12:25 p.m.</em></p>

 

<p><em>Do you mean it is not worth doing at all or that if your intention is to print then it will hardly be noticed?</em></p>

 

</blockquote>

<p>Mike, it's easy enough to do the experiment yourself.<br>

Get a nice still life setup, bracket down 3 stops, all the way through to up three or four stops, to blowout.<br>

Then compare 'em, see where the data falls away from the shadows, see what's optimal, see where you get into trouble on the high side.<br>

It would take me less time to do this than to type it out, and I'm pretty kwik typin' stuff.<br>

It;s a great hands- on learning experience to boot.</p>

<p>Bill P. vs. the Forces of Doom</p>

[ben_goren]

<p>Mike,</p>

 

<p>If I may add on a suggestion to Bill’s, be sure to examine the effect not just on shadow noise;

that’s not in dispute. But also examine the effect on highlight detail. If you can find some dark

styrofoam such as in my example above, that should help give you a feeling for the sorts of situations

where something that, at first glance, appears to be low-contrast dark (and therefore ideal for ETTR) can

be anything but.</p>

 

<p>As I hinted, many surfaces exhibit exactly this sort of texture. Metals, plastics, fabrics….</p>

 

<p>Also, if you use any of the RAW conversion tools for significant post-processing (tone curves, shadow recovery, black point clipping, that sort of thing), be sure to experiment with them, as well. You might be surprised to discover how radically ETTR effects the behavior of the tools you’re used to using.</p>

 

<p>Cheers,</p>

 

<p>b&</p>