expose for shadows, develop for highlights now inverted on digital?

Discussion in 'Digital Darkroom' started by dansutton, Feb 12, 2009.

  1. I'm reading ansel adams' book "The Exposure" and was wondering if the old adage had now been inverted. Adams explains that on film one exposes for shadows and then can use developing techniques to bring back detail that may be over exposed by adjusting developing techniques. Tones can be compressed to pull exposure values of XI or even X down to a more reasonable VII or VIII. and this process doesn't affect the lower valued zones as much, leaving the zones of III and IV more unaffected. Thus expose your shadows correctly and pull down or expand your highlights as needed.
    I'm wondering if this is now essentially inverted. In other words, because of the precipitous shoulder of digital sensors, if it is better to expose the highlight areas better and leave the shadows to fall where they may, and then adjust them in post processing as adams would have done in the darkroom. So for instance, on my Nikon, i could use the d-lighting, or layer masks and exposure compensation for a very targeted approach as well as contrast adjustments. But because of the destruction of detail in overexposing on a digital sensor, if this might be the procedure that would generally yield the most control/best results on digital exposures. Just wondering if any one had a better theory or perhaps saw a fatal flaw in this argument. thanks for your time.
    dan
     
  2. http://www.luminous-landscape.com/tutorials/expose-right.shtml
     
  3. pretty much, yes.. you cannot compare digital capture with film. A sensor is not like film. They are both light sensitive but that's about it.
     
  4. Treat digicams more or less like tranny film. That is, expose for the highlights. In digicams, this is now called expose to the right (that is, push the exposure to the right on the histogram). Just don't clip when you push to the right. ;-)
     
  5. "Treat digicams more or less like tranny film. That is, expose for the highlights. In digicams, this is now called expose to the right (that is, push the exposure to the right on the histogram). Just don't clip when you push to the right. ;-)"
    While I understand this argument I do not understand how to achieve it except perhaps in a few fortunate situations where the exposure curve is JUST right (i.e. flat throughout). We are seldom blessed with perfect exposure curves, in fact at just those times when I worry about getting the exposure correct, it is because the picture is particularly contrasty. In this case there will almost certainly be a tall peak on the right hand side of the curve (and most likely on the left, too.) Which, it seems to me would preclude pushing the curve further right without blowing even more highlights
    So my question is - how on earth does one push the curve further to the right without blowing more highlights in these circumstances? I have to say, my instinct is to do the opposite - deliberately UNDER expose to pull that excess of light pixels at the right hand side of the graph back to something reasonable. In fact I had originally believed that this is what "shooting to the right meant" (i.e. shoot having regard to the need to lower exposure and limit the chance for blown highlights.) But I have read quite a few posts that say the opposite as in the above post .
    PLEASE EXPLAIN BECAUSE I AM NOT GETTING SOMETHING AND OBVIOUSLY THE WORLD KNOWS MORE THAN ME - SOMETHING THATS NOT UNUSUAL , I FIND. :^)
     
  6. SCL

    SCL

    Peter - I have found in my digital shooting the answer isn't always as precise as I'd like. Generally I agree with the concept of exposing for the highlights in digital, like transparencies, but some scenes don't seem to really work out with that approach. My suggestion, since it is digital, is to bracket your exposures and see how your particular gear and style responds, and use your findings as a basis for future shots. After all, it doesn't cost anything more, as it would in film, and you can have your results for analysis same day.
     
  7. Peter, have you read the tutorial that Mark Sirota references? That should explain it for you, both the how and the why. Of course, if the subject brightness range (SBR) exceeds the dynamic range of the sensor, you have no choice but to clip one end or the other (or both) when making your exposure. In that case, make two -- one that preserves the right, one that preserves the left, and combine them into an HDR image. Many recent digicams actually automate this kind of bracketing to make it easy.
     
  8. I believe, Dan, that you have it exactly right. Expose for the highlights and slightly underexpose when in doubt. Which is exactly what you're supposed to do with slide/transparency film, by the way. It is best to use the rules for slide film when shooting digital. Digital could be considered a "positive film", just like slide, because it is recording what is actually in front of the lens, not the negative of it. Because it's not recording the image like a negative film would, the rules are reversed.
     
  9. Basically, in typical situations, expose so that a) the right "point" of your histogram curve does not flatten against the right side of the diplay and b) the blown highlight indicator (flashing black on Canon) does not appear. In portions of the image where the luminosity value goes all the way to maximum the channel can hold no detail at all, and there is essentially no possibility of recovery.
    There are exceptions to every "rule," including this one. In some cases you can intentionally allow some highlights to go completely white as long as that is what you intend. For example, small spectral highlights can sometimes do this.
    While overexposing the highlights on a DSLR may leave you with a complete and unrecoverable loss of highlight details, underexposure of the shadows can be a problem from which some recovery is possible. If your choice is between blowing the highllights and underexposing the shadows, it is probably a better choice in most cases to take care of the highlights at the time of exposure and try to recover the shadows in post. When underexposed your shadows will have two problems: there may be a great deal of noise and there can be some banding of shadow details. Neither of these are ideal, but they are "less bad" in most cases than losing highlights. In many cases a bit of noise in the shadows won't even be noticed in a print, and there are ways to minimize the noise and banding in post.
    When faced with a really wide dynamic range scene you could handle it the way film photographs often did - use a graduated neutral density filter. With digital you also have another option which can sometimes provide even more control: bracket your exposure and combine the two (or more) version with masks in post-production.
    Dan
     
  10. Guys, I have seen the Luminous Landscape article but that is exactly what I am talking about ----look at the first set of curves - the nice rounded curve with a lot of mid tone content. (OK admittedly this would be a little boring as a photo as there would most likely be too much mid tone and too little dark or light contrast) but with a curve like this, why on earth would you want to worry about exposure or moving it to the right - something which with this curve you could in any event readily achieve with much more precision after the event in Photoshop.
    Lets, for argument sake, see the article re-written with a real world curve - spread out with a peak on the right hand side which is what happens when you take a picture (say) on a bright sunny day and have some sky in it or perhaps some shadow which has over influenced exposure producing clipped highlights. Could you move the curve to the right then. I frankly fail to see it. (Sorry I am just trying to understand the argument not be combative.)
    I do get the argument about merging pictures as a HDR image and can thankfully do that easily in Corell Paint Shop Pro Photo X2 as there is an automated HDR filter that will undertake the merge for you. But again this only works in some situations - as when you have had the forethought to carry a tripod so as to get the necessary steady double or triple shots at different exposures and when the image is a static one - not involving moving elements.
    Hmm perhaps I should have gottan that darn fuji S5 instead of the D200!
     
  11. PS to my post above.
    Alternatively does this not mean that my original interpretation of the instruction to reverse the adage about "exposing for shadows and developing for highlights" is correct?
    That is to say does this not really mean when shooting digital, you should let the shadows come out under exposed in favour of having better highlight exposure as in digital its easier to retrieve too dark shadows than it is to retrieve too bright highlights.
    In other words, exposing to the right means pushing your curve further to the left in the knowledge that the situation can be better retrieved in post processing than if the opposite is done?
    This seems much more logical to me than the other argument.
     
  12. ok. now that i have that little conundrum a little more resolved, i have a follow up if any one has a good answer for me. ansel says that contrast is adjusted by the amount of developer added or removed from developing of the negative. and that this disproportionately affects the higher zones: the point being, that, essentially, one can fix one's shadow values, and then telescope or contract the range of the higher zones above that. Now, supposing that i chose a highlight in my photo or the most luminant area that i want to preserve detail in, and expose to have that in zone VII or VIII (very bright and at the threshold of holding detail and texture), and then because of the luminance values of the subject/scene that i am photographing, i find that the shadows in my picture are going to be a little brighter than i would like: i can adjust the contrast, and by increasing the contrast, deepen the shadows. Now, will it also push my highlights higher, and thus i need to compensate for this adjustment, or does it affect the highlights or shadows disproportionately more? The point, i guess, is where is the center of the "stretching"?
     
  13. also, re: G Dan Mitchel: i think that this was exactly what i was thinking. that ansel says blown highlights on film are more recoverable on film whereas underexposed shadows are gone forever on film: the opposite is true on digital sensors and thus setting an optimal exposure for the highlights and then pp on shadows is the ideal way.
    but i'm interested to hear of information about the contrast increase/decrease compensation. For instance, say that i pick my highlights in my scene and want them to reproduce (in black and white) as a bright zone VII or VIII (which is about 80-90% luminance, if i can negligently ballpark it) but either way pretty white. Now for the sake of argument and the example here, let's say that that renders the rest of my scene about neutral gray at 18% and i want to make the scene more dramatic. So i also want to deepen those shadows. So i increase my contrast, say all the way up. That should, at the same exposure settings, deepen those shadows, but will it also then raise the luminance of the highlights on my sensor and screen? do i need to take off a little shutter speed or close my aperture a little to compensate a little bit for this adjustment? on film, it seemed like only the highlights were affected by contrast adjustments.
    thanks a lot for your responses.
     
  14. Guys, I have seen the Luminous Landscape article but that is exactly what I am talking about ----look at the first set of curves - the nice rounded curve with a lot of mid tone content. (OK admittedly this would be a little boring as a photo as there would most likely be too much mid tone and too little dark or light contrast) but with a curve like this, why on earth would you want to worry about exposure or moving it to the right - something which with this curve you could in any event readily achieve with much more precision after the event in Photoshop.​
    Peter.... The concept of 'expose to the right' is this: You want to capture the brightest highlight that you want to retain detail in (so, basically excluding specular highlights) as close to the right hand side of the histogram as possible. You don't want to clip (that is, overexpose) anything important off the right side of the histogram. So in a 'normal' contrast scene, you want to expose say a face or whatever at the right edge of the histogram, but without clipping it. In post-production, you bring the scene back to a 'correct' exposure. i.e. the faces won't be bright highlights, but they will fall in the histogram where they would have had you not exposed to the right in the first place. So, you end up with the same image exposure that you would have had you normally exposed, but your shadows will have far less noise in them.
    Now, in the case of a high contrast scene, it is likely that the scene dynamic range will exceed your sensor dynamic range. So you are correct in asserting that you won't be able to push this exposure any further to the right. The only choice you get in this situation is whether to clip the highlights or clip your shadows.
    Hope that made sense.
    Now, supposing that i chose a highlight in my photo or the most luminant area that i want to preserve detail in, and expose to have that in zone VII or VIII (very bright and at the threshold of holding detail and texture), and then because of the luminance values of the subject/scene that i am photographing, i find that the shadows in my picture are going to be a little brighter than i would like: i can adjust the contrast, and by increasing the contrast, deepen the shadows. Now, will it also push my highlights higher, and thus i need to compensate for this adjustment, or does it affect the highlights or shadows disproportionately more?​
    Dan.... You can increase contrast in whatever fashion you like by drawing the appropriate curve yourself in photoshop. That is, leave the highlight to midpoint region of the curve linear, and pull down the shadow to midpoint region. Using lightroom makes this process really easy, but it is not too difficult to achieve in photoshop as well. Cheers.
     
  15. ahhh bernie. thank you very much. you've taken a lot of mystery out of curves for me. i always blundered through them but now have a more tangible grasp of them. thanks very much.
     
  16. With film the issue is lack of density: shadows with neg film and highlights with positive film. Conceivably you can burn through too much density, but there's no recovering detail that's not on the film. With digital you have to worry about both ends of the scale.
     
  17. Expose to the right:
    http://www.luminous-landscape.com/tutorials/expose-right.shtml
    http://www.digitalphotopro.com/technique/camera-technique/exposing-for-raw.html
     
  18. All well and good but my point is still this. The Luminous Landscape article seems to me to be disingenuous. With a tonal curve that looks like the one it is depicting, exposing to the right is easy. But why on earth would you want to bother in that situation? Neither the shadows nor the highlights are clipped. Noise at the shadow end should not be a problem with a curve like this and neither should clipped highlights.
     
  19. I hear you Peter. I went back to shooting film 90% of the time because I got tired of dealing with these issues. It takes me longer to clean up noise and get nice contrast from my DSLR than it does to scan a 6x7 transparency which I end up doing little to no post-processing on.
     
  20. Exposing to the right is okay, until you clip the highlights. Blank whites are impossible to salvage in jpeg. Clipped shadows are more salvageable. Of course, a full range of values without clipping on either end is the ideal. But if pushed in broad lighting I'd rather have noisy pulled up shadows than totally blown highlights. A better solution of course, is shooting RAW. Then proper exposure is somewhat less critical.
     
  21. All well and good but my point is still this. The Luminous Landscape article seems to me to be disingenuous. With a tonal curve that looks like the one it is depicting, exposing to the right is easy. But why on earth would you want to bother in that situation?​
    Peter... I've just had a look at the curve (didn't look at it earlier), and you are right that in that case there wouldn't be much to be gained in exposing to the right. But there also wouldn't be anything to be lost. And you would actually gain a little bit in reduced noise in the darkest part of the image, whether that is shadows or low mid-tones. But your point is taken.
    One extra thing to be said about that histogram, is that it is a gamma converted histogram (ie. it is from a jpg or tiff etc). The real raw data histogram as captured by the sensor, would be stacked up at the left hand end like nobody's business. So while most of those midtones end up in the middle of the histogram, they actually started out down at the low end. So there is still a benefit to be gained with this technique. But as will all these things, there are upsides and downsides, and ultimately one has to take this into account to determine what is the best approach.
     
  22. Dan, I don't know about the digital stuff to the extent y'all are discussing it here; but, Adams did not always overexpose and underdevelop. There were situations in which he did the opposite; although, if I remember right, one of his more frequent patterns was to underexpose and develop normally. (N-1 to N) or its converse (N to N-1).
    I mention this because it occurred to me that it may not be the recording media, but your choices in response to the situation that was leading you in that direction. Maybe it would be helpful to look at the times when Adams decided that way, too. In "The Negative" he had several ways of doing it.
    If I follow the general concept of the histograms thing right, above; Adams, too, would use compression or expansion of contrast when he was tinkering like this. I think part of the discussion of curves, above, is a computer mimic of this; seems like same idea, but different media. Anyway, "The Negative" by Adams will have more detailed examples.
    If you are short on time, maybe the way to find something that looks like your question would be to cruise through the pictures; there were several in my copy that had notations about exposure and darkroom pushing or pulling for compression or expansion (like N-1 to N or N+1 to N-1, where N is normal development and the pluses and minuses represent adding or subtracting the equivalent of a stop's worth of light, but in the form of chemical exposure in the darkroom and not just in light exposure in the field).
    He frequently recommended Overexpose/underdevelop model, because it is "safer" for capturing, but there were instances in which he noted the opposite. The driving factors were arbitrary and situational; main idea was that he would pick his subject and then push or pull everything else in order to render his subject in the range of grays he desired. His book explains his method better than I can. "The Negative." I'm sure that other one is a good book, too; but, I know there is a wider variety of examples in "The Negative." One of them will answer your questions.
     
  23. ahhh bernie. thank you very much. you've taken a lot of mystery out of curves for me. i always blundered through them but now have a more tangible grasp of them. thanks very much.​
    If curves look odd simple levels adjustment is very obvious as you move the sliders under the histogram.
    I usually just shoot RAW about one stop over jpg-limit. Recover highlights and there's a +1 bonus in the shadows without even touching them. Actually this is a "proper" exposure when you have post editing in mind - not that different from zone system. Shoot for post work, overexpose and underdevelop. Or perhaps I'm just a bit simple. ;)

    Also a matter of taste to some degree, I really don't like underexposing and adjusting muddy images.
     
  24. I've always thought shooting RAW similar to shooting negative film. You can overexpose to get good amount of noiseless shadow detail. *More light is always better* if possible within medium boundaries. Also, RAW is your starting point from which you can make different interpretations. That's it.
    Of course negative can tolerate pretty massive amount of overexposure before totally blowing out and with b&w you can adjust development but this only changes the tolerance, not the principle. Extreme shadow/highlight recovery is not that fun in either medium (I do have a darkroom).
    There's no difference in making an edited JPG and wet printing a negative. It's just that work order is different. In the darkroom you edit while you print and tone but with digital you have already edited and you're left with the problem of calibrating everything so that the print process is neutral. And that's really fun isn't it. :)
     
  25. Peter,
    With a tonal curve that looks like the one it is depicting, exposing to the right is easy. But why on earth would you want to bother in that situation?
    The theory is that exposing to the right reduces noise and improves tonality. In my experience you don't necessarily see or realize this unless and until you push an image hard in post processing.
    Also: some of the confusion about exposing to the right could be eliminated by describing it in terms of a spot meter. Spot meter the brightest highlights you want to hold and place those near the right of the histogram. Let the remaining tones fall where they may. If you're using another metering mode your image may already be "to the right", or even horribly blown out, depending on what the meter decided.
    Incidentally, I am a huge proponent of learning to use your partial/spot meter and relying on it whenever time permits. A lot of exposure problems disappear once you turn off dumb 'smart' metering patterns. Partial/spot also forces you to think about where the tones will fall in the scene and make intelligent choices about it.
     
  26. to John O'Keefe-Odom. Yes, he did master his camera and manipulate it to represent the picture as he wanteed it. truly amazing hearing him explain his thought process for exposures. Yes, the confusion we might have as well is that what we call over-exposure he called more exposure and less exposure; he called "over-exposure" an error in exposure. and i think that you bring up a good point: photography is not, and should not be, formulaic. there should be no algorithm to taking pictures; what adams did was picture a visualization in his head and then know his craft so well that he knew how to reproduce the vision in his head as a grey scale image. the point being that he never followed a formula for photography, but, armed with a goal, knew what steps to take to achieve that end.
     
  27. The Luminous Landscape article seems to me to be disingenuous. With a tonal curve that looks like the one it is depicting, exposing to the right is easy. But why on earth would you want to bother in that situation?
    I've always thought shooting RAW similar to shooting negative film.​
    Gang, film isn't linear! Anyone remember H&D curves? Digital capture is linear. HALF of all the data of your capture is contained in the first stop of highlight data, the least in the last stop of shadow. Its simple! You expose to the right to get as much data (and as little noise) in that last stop where all the noise resides. Got nothing to do with tone curves at this point. If you expose for a gamma corrected JPEG, look at the big fat lie in terms of the histogram or color on the camera LCD, you're fooling yourself in terms of the Raw data exposure. You want to under expose your film or digital, go ahead. ETTR isn't about over exposure, its about optimal exposure for digital linear capture.
    Todays meters are built for film too. Figure the true ISO sensitivity of your chip. Figure out what is the proper exposure shy of blowing out highlight data you wish to render. That's the correct exposure.
     
  28. Exposing to the right is okay, until you clip the highlights.​
    That's not ETTR, that's over exposure. And of course, the opposite is under exposure <g>
    Digital is no different then film in terms of deciding what is the dynamic range of the scene (here spot meters are useful). Doesn't matter if its you, me or Adams, we have to decide how to fit what is more often a far larger exposure latitude then we can capture. What is the specular highlight in the scene? Where do you want to place the clip under which you decide what white deserves detail? If you want detail and you blow it out, you didn't expose properly. Doesn't matter if its film or digital. The difference is, our meters and cameras are working as if we were exposing a non linear capture and that's simply not happening. Expose for the highlight, period, meaning, don't blow out data you want to record. ETTR never has suggested otherwise.
     
  29. This a thread that I wish to follow.
     
  30. As an experienced slide shooter and traditional darkroom printer, I am in agreement with Bruce and Jennifer: ETTR in digital is very similar to exposing for the highlight on slides.
    In either a traditional or a digital workflow of creating images, there are two separate stages - getting the best exposure for the desired effects in camera, and post processing in the darkroom for the final images, for web or in prints. Many photographers, such as AA, would expose an image in a certain way with the anticipation of post processing it accordingly later in mind. Between the traditional and digital workflows, the techniques to achieve these goals in each stage have changed somewhat. But the fundamental separation between the two stages have not. Some of the questions seem to be confused by the two, or treating the two in one lump.
     
  31. Just to clarify something that was said way above:
    You don't want to clip (that is, overexpose) anything important off the right side of the histogram.​
    That is not quite right. You do indeed want to overexpose, if that is what it takes. You just don't want to clip. You can fix the overexposure in PP, and that will result in a higher overall S/N ratio, you can't fix the clipped portions.
     
  32. That is not quite right. You do indeed want to overexpose, if that is what it takes. You just don't want to clip. You can fix the overexposure in PP, and that will result in a higher overall S/N ratio, you can't fix the clipped portions.​
    How so? Clipping is the case where all three resulting channels are 255/255/255 (using an 8-bit encoding) no matter the rendering settings. Clipping data you want IS over over exposure. Exposing to retain that data is proper ETTR. Its real, real simple. ETTR=No clipping of data you don't want clipped. Clipping data you wish to retain is over exposure.

    You're not fixing anything. You're setting the proper rendering controls based on the exposure. There's nothing broken to fix. You can't fix true clipping, you need to expose properly for that data.

    Note this is ALL based on the rendering settings and histogram feedback in the converter NOT the camera! The histogram and clipping is only based on the JPEG you'd get if you didn't shoot Raw. Everything provided on the back of the LCD is science fiction in terms of the Raw data.

    Note that in Adobe Raw converters, there's a somewhat unique function whereby if one of the three color channels has some data, one channel is not fully clipped, the other two can be recreated using that one channel (this is what Recovery or proper curves use provides). In most other converters, if any channel clips, it treats all channels as clipped.

    In terms of over exposure and proper testing, on my older 5D, I could "over expose" 1.5 stops over what the incident meter suggested was indeed correct exposure and still retain all highlight data (on a very, white subject) by altering Exposure in LR/ACR. That was the correct ETTR, the exposure suggested by the meter was down 1.5 stops. 2 stops over, there was no data to pull back (all three channels clipped). Its all shown in the article URL above from Digital Photo Pro.
    *babelcolor.com/main_level/White_Target.htm This is an ideal object to test for true clipping.
     
  33. Well folks just when we need Ansel Adams to solve this sticky problem, he is gone.......Maybe we will just have to finally work this out ourselves........I think has left us a few clues though........Jim
     
  34. The last several posts here are debating semantics rather than the core issue, so I will (foolishly?) attempt to clarify:
    For some, over-exposure means more exposure than the meter recommends .
    For others, over-exposure means enough light to clip the highlights .
    Let's be careful about that term and be clear what we mean when we use it; don't assume that others have the meaning in mind.
     
  35. Yes, those are two significant differences regarding over-exposure.
    But Andrew's note about digital capture being linear is very important to remember.
     
  36. I think Andrew pretty much has it right...just read his article in the Digital Photo Pro website. Pretty much understand it, have been doing a lot of it on my own for a few years...more by trial and error, than his calculated method. But his analysis of how to custom render the ETTR RAW image in LR or photoshop is something I really did by the seat of my pants...will have to analyse my stuff via his findings....same cam, so it should work. I just did it all by eye to bring the image back to what I wanted it.
    I think if everyone just took a few hours to read the article and follow th steps he lays out, you'de all see the benefit of ETTR.
    And by the way, for the umpteenth time, it is NOT overexposure, its correct exposure. You expose for the highlites in the camera, and you develop (ie adjust in RAW) for the shadows (and everything else, actually)
     
  37. For some, over-exposure means more exposure than the meter recommends .​
    The meter is wrong! Its again, very simple. If you set the meter to ISO 200 and shoot at ISO 400, doesn't matter a lick what the meter says is "correct" or "over exposed" its the wrong exposure.
    Point reflective that meter at a snow bank and tell me if you get the "right" exposure. It will be grossly under exposed for digital and still under exposed for film.
    Over exposure is too much exposure period. Clipping is over exposure assuming you don't want to clip those areas. Simple.
    And by the way, for the umpteenth time, it is NOT overexposure, its correct exposure.​
    Amen to that!
     
  38. I would also volunteer that clipping overexposure is not when all channels max out (255,255,255). It is if ANY of the three channels max out. If any channel clips, color accuracy will suffer, sometimes dramatically. In a typical RAW processing scenario, you try to 'recover' some cloud detail only to find strange colors as you do it--that's because you've reached the point where one or more channels clipped. This kind of thing has bit me more than once when I was trying to expose to the right but didn't take into account that one channel was going too far. I now like to use the multi-channel histogram review if available.
    I think another area that could use some clarification is that in terms of capture, we might say that any undesireable clipping is overexposure. When it comes time to create final output during post-processing, then the 'exposure' level is not so much a technical limitation but an artistic interpretation--a dark room exposed to look like daylight might be considered overexposed even though there are no blown highlights.
     
  39. I would also volunteer that clipping overexposure is not when all channels max out (255,255,255). It is if ANY of the three channels max out.
    At the point the sensor is hit with the photons, there are not three channels, a Raw file is essentially a single channel grayscale file with a set of filters used to eventually produce those three channels when demosaicing in the Raw converter (or camera) takes place.
    In terms of defining exposure, I'm sure we can all search the web, but I don't see where Wikipedia isn't serving the definition of correct exposure or clipping correctly:
    http://en.wikipedia.org/wiki/Exposure_(photography)
    Exposure settings
    "Correct" exposure may be defined as an exposure that achieves the effect the photographer intended[3]. A photograph may be described as overexposed when the subject tones seem too light, or underexposed if the subject seems too dark. Underexposure may cause loss of shadow detail, also known as crushed blacks, and overexposure may cause loss of highlight detail, also known as blown-highlights.
    The purpose of exposure adjustment is to control the amount of light reflected (or emitted) from the subject so that it conforms to the the film (or sensor's) exposure sensitivity.
     
  40. Mark,
    The last several posts here are debating semantics rather than the core issue, so I will (foolishly?) attempt to clarify:
    For some, over-exposure means more exposure than the meter recommends .
    For others, over-exposure means enough light to clip the highlights .
    Bingo. You captured exactly what I was trying to illustrate. There's confusion in this thread because people are taking their meter recommendations as 'correct', and therefore thinking that ETTR is 'overexposure'.
    Multi-segment meters are not 'correct', they represent a best guess by a machine. A guess biased to producing pleasing results when going straight from file to print. ETTR is for people who treat the file as separate from the print, something which must be manually processed into the print. For these people ETTR is a natural consequence of the desire to capture as much information as possible knowing that the scene values can be mapped to the desired tone values for a print.
    Andrew,
    Over exposure is too much exposure period. Clipping is over exposure assuming you don't want to clip those areas. Simple.
    Only if you're going straight to print from the file. If you're going to post process then overexposure = clipping highlights you wanted to hold. ETTR, for the person who will post process the file into a print, is not overexposure. It's just a simple guidline for capturing the maximum amount of scene information.
     
  41. Over exposure is too much exposure period. Clipping is over exposure assuming you don't want to clip those areas. Simple.
    Only if you're going straight to print from the file.​
    Sorry, I have no idea what you're talking about. You can't take a Raw file directly to print. Its not even an image at this point. You've got to render it, demosic the data, fit the dynamic range to the print etc. You can't view, print or deal with Raw data until this is accomplished (http://en.wikipedia.org/wiki/Demosaicing).
    Read: http://wwwimages.adobe.com/www.adobe.com/products/photoshop/family/prophotographer/pdfs/pscs3_renderprint.pdf
     
  42. Andrew,
    I had read the LL article and now just read your DPP article. I hadn't realized the in-camera histogram was based on the in-camera rendered JPG (should have been obvious) and therefore is problematic when trying to gauge the RAW data - thanks for pointing this out! It seems I need to do the work and figure out my ETTR technique in greater detail.
    I wanted to point out one thing though. You wrote "it would be useful if our cameras could bracket in one direction: from “normal” to increased exposure in a preset number of ƒ-stop increments." I don't know what camera you are using, but my entry-level Rebel XS (1000D) can do this easily (well, it brackets 3 exposures, I don't know if I can do more than 3). I simply set the bracketing in the menu (which brackets "around" the metered 0 level), and then back at the shooting screen spin the control dial while holding exposure compensation to "shift" the bracketing to the right. Bingo - I get bracketed shots at [0, +1, +2]. Or whatever stop intervals you like.
    Edit: I just tried this to make sure I wasn't misremembering the exact button combinations, and discovered I can use this method to bracket up to [0, +2, +4] (!). That's more "exposure compensation" than you can normally get from a single shot (unless, of course, you shoot full manual). And naturally it doesn't have to start at 0; I was able to set it to [+1&1/3, +2, +2&2/3] and all sorts of other funky permutations.
     
  43. ETTR? Exposure to the right. Took me ages.
     
  44. Andrew,

    Sorry, I have no idea what you're talking about.

    You said: Over exposure is too much exposure period.

    I pointed out that this only makes sense if one is going straight from file to print. Sarcasm about RAW conversion doesn't change that (I don't need a link to explain RAW, thank you very much). You can convert/print a RAW file without playing with the levels, taking the camera exposure pretty much as is.

    But if you normally do play with levels and tone values, then you've only overexposed when you clip a highlight you wanted to keep.

    In short, I disagree with your definition posted earlier.
     
  45. Edit: I just tried this to make sure I wasn't misremembering the exact button combinations, and discovered I can use this method to bracket up to [0, +2, +4] (!).​
    Yes indeed. Another reader pointed this out and showed me how to do this on my old 5D using custom functions.
     
  46. You can convert/print a RAW file without playing with the levels, taking the camera exposure pretty much as is.​
    OK, I see now what you're saying although it seems kind of silly. So take the Raw "as is" from the default converter and print with no additional rendering settings? I don't know why you'd do that. And in fact, we don't know if the data is exposed correctly or not (more than likely, not, as ETTR requires you build a "Normailzie" setting so it doesn't "look" over exposed. And its not over exposed IF you set the rendering and don't blow out the highlights you wished to retain.
    The Raw data, despite the appearance in the converter at any time, may or may not be over exposed. If you have highlight data you hope to retain and you can't retain it no matter the settings, you over exposed!
     
  47. Let me try again, and this time I will state exactly what I mean:
    Correct exposure - The set of settings which will reproduce the scene exactly as one would want the final image to appear, whatever that may be.
    Underexposure - The set of settings which will reproduce the scene darker than one would want the final image to appear.
    Overexposure - The set of settings which will reproduce the scene lighter than one would want.
    Note that none of the above speaks to clipping.
    Clipping - The result of allowing any one of the color channels to saturate at its maximum value.
    So, I think it should be clear that will these definitions, and given the reality of noise as a fixed percentage of full scale capability for a given set of conditions (camera, gain (ISO), exposure, temperature) in most cases what should be done is to simply avoid clipping to the degree that makes the most sense, which may include (according to my definition) overexposure.
    To me this is what ETTR means. I was simply attempting to distinguish between clipping and overexposing just as Mark said.
     
  48. Your definitions are fine. My problem is losing clipping because it truly describes what's happening to our data. There is none that might have been had I exposed differently.
    Overexposure - The set of settings which will reproduce the scene lighter than one would want. But where in the tone curve and can we darken this so it is desirable? If the data is clipped, the answer is no.
    Clipping is important in this discussion of ETTR and exposure because if you are not careful, you go too far and clip the data you didn't want to clip. It defines the problem and exactly where the problem occurs and its a problem that can't be fixed with image processing. Clipping is very specific and over exposure is pretty ambiguous. And when trying to gauge the proper ETTR settings, the only way I know of is to place a very white object like the BableColor tile, bracket and see when it clips based on the controls I have in my Raw processor. There's a limit to what I can recover using the provided tools, anything that results in no tonal data is clipped and thus over exposed assuming I didn't want it clipped at capture.
     
  49. Below is an example of an image I forgot to ETTR shooting hand held in the shade. You'll note the plots on the curve pretty much squeezed all the detail in a tight spot, but the main point of this sample is to show the distribution ratio of available tones to work with when trying to edit such a file even in 16 bit ProPhoto RGB and the unavoidable posterization that occurs. I even faked ETTR by normalizing (lightening) the image to make the histogram resemble my camera's as if I had ETTR'ed and then saved in 16 bit tiff and edited in CS3. Still got the posterization.
    The original full image on top was arrived at with minimum editing because it happened to already resemble the scene as I saw it. All that I applied was +.70 Exposure, +63 Contrast and a color temp adjust. Even though this shot looks done, it's a very terrible image to apply extreme edits on as shown in the ACR screengrab.
    00SSIu-109793584.jpg
     
  50. To clarify these fine definitions, and at the risk of adding confusion, I have added some comments.
    Correct exposure - The set of settings which will reproduce the scene exactly as one would want the final image to appear, whatever that may be.
    A correct exposure is what a photographer subjectively chooses for an image. There are no objective definitions for settings to arrive at a correct exposure. In this context, a correct exposure in camera may mean no clipping, or clipping highlight, or clipping shadow, or clipping both. All are done intentionally depending on how a photographer wants the final image to look like.
    Underexposure - The set of settings which will reproduce the scene darker than one would want the final image to appear.
    Overexposure - The set of settings which will reproduce the scene lighter than one would want.
    When a viewer describes an image as under- or overexposed, he/she means that the image looks too dark or too light in his/her eyes, subjectively. There are no objective definitions for under- or overexposure. An under- or overexposed image may or may not have any clippings.
    Clipping - The result of allowing any one of the color channels to saturate at its maximum value.
    Avoiding in camera clipping may or may not result in an image a photographer desires. Conversely, he/she may intentionally include clipping to be part of a desired image.
     
  51. IMO, a correct exposure doesn't clip highlight or shadow, within the limits of the film/camera. If a photographer chooses to overexpose, or to underexpose, that is a personal choice, but it does not equate to a correct exposure. JMO
     
  52. IMO, a correct exposure doesn't clip highlight or shadow, within the limits of the film/camera.​
    The only thing I'd ad to the above is taking the scene dynamic range into account. Say the scene has a 15 stop DR, I'd like to clip neither but my capture device can only capture 13 stops. In this case, short of HDR/multiple captures, we have to decide where to clip (shadows to retain highlights, or highlights to retain shadows). But every photographers has struggled with this since day one. So:
    A correct exposure doesn't clip highlight or shadow, within the limits of the film/camera/dynamic range of the scene.
     
  53. That's what I meant, but didn't spell out, Andrew. DR is the limiting factor for sure.
     
  54. A correct exposure doesn't clip highlight or shadow, within the limits of the film/camera/dynamic range of the scene.​
    Andrew,
    Not to belabor the point but I would add to this statement: ...And exposes those highlights (in which one wishes to preserve detail) sufficiently that they just approach the point of clipping.
    That, I think, is the essence of ETTR.
     
  55. Not to belabor the point but I would add to this statement: ...And exposes those highlights (in which one wishes to preserve detail) sufficiently that they just approach the point of clipping.
    That, I think, is the essence of ETTR.​
    Agreed!
     
  56. Look into HDR imaging. Shoot raw, and get the book "Mastering HDR photography by Mike Freeman. You can't use it for sports and stuff, but many of the users will find it helpful.
     
  57. jtk

    jtk

    Lightroom seems almost magical, recovering highlight detail that appears to be lost...infinitely more efficient than Photoshop for the purpose.
     

Share This Page