PetKal wrote:
That is all true, Anders my friend, if a key photography objective was to have the least amount of noise in an image. However, that is often only an auxilliary technical aspect.
I often "expose to the left" where that suits my photographic quality objective better. True enough, here and there I thereby create pronounced noise which has to be dealt with, yet I gain much more in my own aesthetic sense.
For example, I must have seen thousands of images of birds with white colours in their plumage which have been exposed to the right following the internet mantra. Now, as a result, the pictures might exibit some combination of these three adverse appearances:
(1) Highlights are unpleasantly blown in the image, and details are gone where and when it counts.
(2) After a major photoshop intervention, the blown highlights have been "mitigated" in a way that turns them into grey-beige, which is clearly not the target's colour.
(3) The dark/irridescent plumage is left overexposed, I guess in order to "bring out the colour", which again departs from the true target's colouring and renders an unnatural look to the bird....Show more →
ETTR isn't just about getting lower noise but getting better tonal separation, especially in the shadows. This ends up being most important in landscape photography.
#1 and #2 are simply incompetence with exposure and have nothing to do with ETTR. As for #3, it sounds more like a post processing decision, one that you disagree with but that others may consider being done for artistic reasons and again I would argue don't have anything to do with the SNR maximizing principle behind ETTR.
You know what? In real photographic terms - say as observed in a 20" x 30" print - choosing the "native ISO" doesn't make a darn bit of difference.
Avoid the extended ISOs (like ISO 50). Generally use the lowest ISO that seems reasonable give the parameters within which you are shooting. You'll be fine.
I shoot a 5D2. If you gave me two versions of one of my better photographs, one shot at ISO 100 and the other shot at ISO 200, I couldn't tell you which was which.
Dan
winman3 wrote:
I see references to a particular camera's "native ISO" but no one ever mentions this mystical ISO.
So, under favourable conditions, i.e. f-stop and aperture can be widely adjusted, movement not a factor, etc. etc. is there a "BEST" ISO to use for a particular camera? Or just use the lowest ISO setting for optimal results?
thedigitalbean wrote:
ETTR isn't just about getting lower noise but getting better tonal separation, especially in the shadows. This ends up being most important in landscape photography.
#1 and #2 are simply incompetence with exposure and have nothing to do with ETTR. As for #3, it sounds more like a post processing decision, one that you disagree with but that others may consider being done for artistic reasons and again I would argue don't have anything to do with the SNR maximizing principle behind ETTR.
Expose to the right as much as needed (in RAW or JPEG) in order to preserve any highlights which are desired in the final print or target medium.
I realize that there will always be exceptions and compromises but, in general, is this an axiom we can use?
I think Anders, Dan and others are saying much the same thing yet looked upon from different angles.
When doing real photography, "expose to the right (or left, or center)" should not be a governing principle of the process. There is a host of other issues which take precedence in the creation of a photograph. However, one should perhaps be cognizant of the different noise and tonal impact which over- or under-exposure has on the final product.
In applying such awareness, we do it differently because we all have different objectives and aesthetic criteria.
Let me remind you of the exquisite "Egrets" image made and shown on this board by Hrow (Henry). That one is just about the most extremely "exposed to the left" image you have probably seen, yet it works beautifully because it was driven by Henry's vision, and not by shadow noise considerations and some such pixel peeping minutiae.
Noise is a given with any scene that exceeds the sensor range. Lower ISO is simply one strategy to minimize it. The better strategy is to eliminate it by finding ways to fit as much of the scene range onto the sensor as possible. That is possible indoors or outdoors in the foreground if using two flashes in an overlapping key-over-fill configuration.
A single flash used outdoors off axis can't lift the shadows it doesn't hit, which is why the second is needed for fill to lift the shadows beyond the level of skylight fill. The best strategy to eliminate noise is to overfill a bit to make the shadows lighter than normal, increasing the S/N ratio there, then darken them back down in PP.
When flash can't be used outdoors lighting angle can be used to reduce scene contrast and put more of the important detail in the "key" light of the sun (i.e. sun at your back) to avoid having important detail fall into the shadow noise range. That doesn't eliminate noisy shadows it simply allow you to eliminate the noise with NR in post-processing without the dark void shadows detracting from the understanding and reaction to the content.
That sun over your shoulder strategy works OK for action shots of people, such as sports photojournalism, but not for portraits due to the harshness of the sunlight and squinting resulting when the sun as "key" light is put in the eyes. That's when it helps to have a couple powerful flashes in your bag.
I keep a flash on a bracket on my camera at all times outdoors and look for opportunities where I can shoot into the shadows of the ambient with something interesting in the foreground which makes the unavoidable lack of detail in the sunny background go unnoticed...
Even in situations like below when the flash is too far away to correctly expose the shaded foreground any light it adds will help raise the highlight and midtone content in the shade above the noise threshold. http://super.nova.org/TP/FillOutdoors.jpg
Every little bit of signal able the noise threshold help when manipulating the image in post processing.
I use a three pronged approach: 1) I use ambient light angle and flash to fit scene to sensor to the extent possible at capture, 2) I try to put up the middle- and 3/4-tones in the RAW file with ACR, 3) I adjust the tonality locally in CS5 with masked adjustment layers.
The second step tends to amplify noise. For example heres a backlit shot exposed for highlight detail... http://super.nova.org/TP/HSS/_MG_5034.jpg
Shaded side is 3 stops darker (Sunny 16 / Shady 5.6) so on the 6+ stop sensor the white shade towel is rendered middle gray (with a blue bias from the skylight) and the 18% card is rendered nearly black. But in ACR the midtones can be lightened to the point of looking normal... http://super.nova.org/TP/HSS/_MG_5034_ACRedit.jpg
but that amplifies the noise in the areas below the sensor noise threshold at capture.
Masked adjustment layers allow more selective modification. Below I started with the ambient only JPG copy above and applied screen to lighten the card, and darkened the shadows to hide the noise there... http://super.nova.org/TP/HSS/_MG_5034_AdjustLayer.jpg
What I'll typically do in that situation when I can't use flash is not apply as much correction at the ACR stage as in the extreme example above, which lightens the midtones some without exacerbating the shadow noise, then complete the "normalization" process with the adjustment layers.
I don't as a rule use any noise reduction because it globally affects the file in ways I can control precisely. Since the shadows with a lot of noise don't have any recorded detail worth saving I just make them darker in PP so the noise isn't seen. A different path to the same goal of making the photo a realistic facsimile of how the scene would be perceived in person.
PetKal wrote:
I think Anders, Dan and others are saying much the same thing yet looked upon from different angles.
When doing real photography, "expose to the right (or left, or center)" should not be a governing principle of the process. There is a host of other issues which take precedence in the creation of a photograph. However, one should perhaps be cognizant of the different noise and tonal impact which over- or under-exposure has on the final product.
In applying such awareness, we do it differently because we all have different objectives and aesthetic criteria.
Let me remind you of the exquisite "Egrets" image made and shown on this board by Hrow (Henry). That one is just about the most extremely "exposed to the left" image you have probably seen, yet it works beautifully because it was driven by Henry's vision, and not by shadow noise considerations and some such pixel peeping minutiae. ...Show more →
Again, if the aesthetic requirements of the image demand that the highlights are to be preserved by "exposing to the right," then one must do so (notwithstanding use of any other technical aides, such as fill-in flash, etc.).
If the needed image highlights are overexposed (e.g. "blown out"), then the aesthetics desired in the image cannot be recovered, no matter what techniques are used in post. Lost data is simply that, "lost data."
If shadow detail is needed, then the appropriate exposure to the left could be employed.
A related (broad) topic is maximizing the overall scene DR for the desired image aesthetics.
But I digress. We've already expanded the scope of this thread from the original question posed by the OP.
PetKal wrote:
Let me remind you of the exquisite "Egrets" image made and shown on this board by Hrow (Henry). That one is just about the most extremely "exposed to the left" image you have probably seen, yet it works beautifully because it was driven by Henry's vision, and not by shadow noise considerations and some such pixel peeping minutiae.
That is one image I for sure remember, and one of the few seen on this board that I will never forget. It was all about the vision, still the technical quality did not stand in the way.
Using an "optimal" ISO, such as 100 on a 5D2, is a bit like shooting at, say, f/6.3 with a particular lens because careful testing has determined that at that aperture you get the highest average resolution across the entire frame, and aberrations are minimized. It doesn't hurt to be aware of such things, but in the real world with good lenses and good (i.e., recent) DSLRs, decisions about settings would usually be based primarily on conditions. Is the subject moving? How fast? What is the light like? How much of the scene do we want to be in focus? Are the shadow areas important to the composition? Is highlight detail an important part of the texture that we want to show? I will certainly admit that when I'm shooting a landscape with a tripod, ISO 100 is my starting point, but going nuts deciding whether to set ISO 160 or 200, 200 or 50, etc. is not something I want to bother with. In fact, I have my bodies set to only allow "full stop" ISOs, just for simplicity and speed. (Apertures and shutter speeds I do by third stops. Works for me.) Sometimes for action, I use Manual exposure and auto ISO on the 1D4, and it then sets intermediate ISOs when needed. I never find noise differences jumping out at me. YMMV
I think it is a sign of a well-functioning discussion arena when we can have a healthy blend of technical details, and a broader perspective with common sense.
If the technical questions don't get answers, we may not be able to utilize our gear to the outmost in demanding situations.
If the common sense and broader perspective is lost, the technical details may be blown out of proportions.
The artistic merit of photographs I prefer to discuss long away from FM gear talk, with people who may have never seen a camera.
alundeb wrote:
I think it is a sign of a well-functioning discussion arena when we can have a healthy blend of technical details, and a broader perspective with common sense.
If the technical questions don't get answers, we may not be able to utilize our gear to the outmost in demanding situations.
If the common sense and broader perspective is lost, the technical details may be blown out of proportions.
The artistic merit of photographs I prefer to discuss long away from FM gear talk, with people who may have never seen a camera.
I'll give you an unprocessed image/example shot today for the application of heavy duty "expose to the left" strategy.
JPG SOOC, 50% crop and resize for web.....no pp other than that.
To my eye the exposure is more-or-less exactly where I wont it to be. And histogram is of no use to me, because I can still see the image with my own eyes on the LCD, and I can see the overexposure blink.
Noise in the background ? I do not care. If anything, I might make the background even darker in photoshop.
Optimal ISO is not the same as native ISO. Native ISO is a property of the sensor and is the sensitivity of the sensor when zero gain is applied to the signal. Not all cameras give you access to native ISO. There may always be some amplification applied.
Before "optimal" can be evaluated one must first settle as what you are optimizing (noise, DR, asthetics, etc). They won't all have the same result.
As far as "expose to the left" dogma is concerned what is often missing is that this is NOT the correct final exposure. When exposing to the left for purposes of maximizing DR or reducing shadow noise (arguably the same thing) you should be returning the image to the "correct" exposure in post processing. I put "correct" in quotes because in the art of photography this is defined by the photographer's desired effect. Exposing to the left should give you the best RAW image to begin your post processing journey. If the goal is out of camera perfection then exposure to the left is only going to suit a subset (and a small subset) of conditions.
PetKal wrote:
I'll give you an unprocessed image/example shot today for the application of heavy duty "expose to the left" strategy.
JPG SOOC, 50% crop and resize for web.....no pp other than that.
To my eye the exposure is more-or-less exactly where I wont it to be. And histogram is of no use to me, because I can still see the image with my own eyes on the LCD, and I can see the overexposure blink.
Noise in the background ? I do not care. If anything, I might make the background even darker in photoshop.
PetKal wrote:
I'll give you an unprocessed image/example shot today for the application of heavy duty "expose to the left" strategy.
JPG SOOC, 50% crop and resize for web.....no pp other than that.
To my eye the exposure is more-or-less exactly where I wont it to be. And histogram is of no use to me, because I can still see the image with my own eyes on the LCD, and I can see the overexposure blink.
Noise in the background ? I do not care. If anything, I might make the background even darker in photoshop.
Very nice image!
I too might be inclined to darken the background (slightly), but the subjective aesthetics are very pleasing and I believe the exposure is very much "spot on" with regards to that (at least to my eyes, as viewed on my monitor).
Darkening the background would also provide for more subject isolation...
For me ETTR is all about PP flexibility. Adjustments to the tonal curve and especially to colors can make noise much more apparent, esp. for the blue channel whose exposure can lag the green channel by 1-2 stops. The PP problem is aggravated on Canon sensors due to their CFA designs and their associated color selectivity (or lack thereof). Starting with the highest possible exposure allows me to do those PP adjustments while keeping noise at a minimum.
snapsy wrote:
For me ETTR is all about PP flexibility. Adjustments to the tonal curve and especially to colors can make noise much more apparent, esp. for the blue channel whose exposure can lag the green channel by 1-2 stops. The PP problem is aggravated on Canon sensors due to their CFA designs and its associated color selectivity (or lack thereof). Starting with the highest exposure possible allows me to do those PP adjustments while keeping noise at a minimum.
I can't remember where I read this, but aren't the higher EV components of an image encoded with more discrete "bits" of information coming from the 14-bit ADC?
And the lower EV components of an image are encoded with fewer discrete "bits" of information coming from the 14-bit ADC?
And why "exposing to the right" may help to provide more "effective dynamic range," due to the encoding methods employed in the 14-bit ADC?
This could also partly explain why obtaining great shadow detail can be so difficult and problematic at times?
alundeb wrote:
This is an example of ETTR in practice
I like the image! (And I can see some noise )
Well, Anders, I'd call that ETTL....because by not allowing any significant highlite clipping on the bird (a very localized clipping is OK and sometimes inevitable), I have effectively pushed the entire histogram to the left.
If one is doing a landscape scene with rather uniform luminance and uniform emphasis thru the scene, then the global histogram starts to make some sense.
However, when shooting discrete smaller (%-wise of the whole metering area) targets, I think histogram is as useful as a third tit.
If for example that seagull morphed into a raven, then I'd really have to push exposure to the right like mad, even if I used fill flash.
cameron12x wrote:
I can't remember where I read this, but aren't the higher EV components of an image encoded with more discrete "bits" of information coming from the 14-bit ADC?
And the lower EV components of an image are encoded with fewer discrete "bits" of information coming from the 14-bit ADC?
And why "exposing to the right" may help to provide more "effective dynamic range," due to the encoding methods employed in the 14-bit ADC?
This could also partly explain why obtaining great shadow detail can be so difficult and problematic at times?
You probably read it in the original Luminous Landscape article about the subject (http://www.luminous-landscape.com/tutorials/expose-right.shtml). Image sensors are generally linear devices whose bit depth increases by a power of two for each brighter stop of light. This extra bit depth has the potential to help the tonality of the tones for a given stop of light (gradations of light) but doesn't impact the dynamic range directly. Experiments performed on some dpreview threads demonstrate that the bit depth for these higher stops is already more than is necessary so you don't really gain anything for ETTR in this regard. ETTR reduces noise by increasing the amount of light captured, not by increasing the number of bits used to describe that capture (as was believed by some at the time the original LR article came out).
Shadows are noisy not because of their reduced bit depth on the tonal scale in a linear sensor but because they have less light, and less light means more noise. In fact it is because the shadows are inherently noisier that they don't need as many bits to describe their tonality (the extra bits would be wasted on describing the noise rather than the tones).
ETTR doesn't increase dynamic range. It just increases the IQ of the dynamic range you do capture. If anything ETTR can effectively reduce the dynamic range if your exposure decision involves clipping highlights which are unimportant to you.
PetKal wrote:
[Well, Anders, I'd call that ETTL....because by not allowing any significant highlite clipping on the bird (a very localized clipping is OK and sometimes inevitable), I have effectively pushed the entire histogram to the left.
Suddenly I can see the light. This is about semantics.
The term "Expose To The Right" is totally misleading. What I mean, when I say "Expose To The Right", is "Expose From The Right". And I am sure I am not alone. We tie up the brightest part of the image (The right hand end of the histogram) at the clipping level. The direction of where you move things may either be To The Left or To The Right, but we start at the right hand side, hence "EFTR".
I am sure you masters of the English language can refine my idea about a renewal of the ETTR to EFTR.
alundeb wrote:
Suddenly I can see the light. This is about semantics.
The term "Expose To The Right" is totally misleading. What I mean, when I say "Expose To The Right", is "Expose From The Right". And I am sure I am not alone. We tie up the brightest part of the image (The right hand end of the histogram) at the clipping level. The direction of where you move things may either be To The Left or To The Right, but we start at the right hand side, hence "EFTR".
I am sure you masters of the English language can refine my idea about a renewal of the ETTR to EFTR. ...Show more →
There you go, we are looking at different sides of the same fence.
Here is another example of a darker bird with larger environmental contents in the frame.......the exposure is still pushed to the left, in terms of the histogram. However, if we are talking a more practical and meaningful language of photographers, one could perhaps say that a degree of underexposure has been chosen for this image (on purpose).
