A Hoodie and an Exposure Lesson/Discussion

[Doug Brown]

I photographed this Hoodie the other day, and my exposure choice led to a discussion with Miguel about the advantages and disadvantages of ETTR (expose to the right). I'm not a big fan of ETTR. When you push to the right, you risk unrecoverable overexposure. Letting in more light forces you to scrub shutter speed and/or DOF. People don't expect to see crazy amounts of details in the darkest blacks; a little detail suffices. And that's easy to accomplish in post processing, as I'll demonstrate below. I am solely concerned with getting the whites right in camera. First the image.

Canon 1Dx, 400mm f/5.6, f/5.6, 1/2500, ISO 640, manual exposure, hand held

[Doug Brown]

And now for a side-by-side of the unprocessed (left) and LR-processed (right) RAW file. I raised the shadows +26, raised the blacks +23, and pulled the whites down -17. I also used an adjustment brush to slightly bring up the blacks and tone down the brightest whites.

I'm interested in your thoughts on my exposure choice, and also on how you would handle the exposure (with an explanation of why).

[Mike Milicia]

Beautiful portrait, Doug!

Here's my two cents with respect to your exposure discussion :

I am a big fan of ETTR and I would say that you applied the concept perfectly for this particular capture to maximize the quality of the image … let me explain …

ETTR is a key element in minimizing noise in an image. The idea is to let in as much light as possible during capture time without overexposing any highlights that you care about. But if part of your subject is bright white and you simply set the correct exposure, there will be very little, if any, headroom to play with so ETTR cannot be applied to great effect and extreme caution must be used to avoid overexposing the whites to the point where they cannot be recovered. On the other hand, if the brightest thing in the scene is a mid-tone, ETTR can easily be applied and will greatly reduce noise in the image. You can overexpose such a scene by as much as 2 stops and not lose any information since the brightest thing in the scene was a midtone. If you do this and then bring the exposure back down 2 stops in RAW conversion, you will end up with the same image but with far less noise than if you had exposed it "correctly" at capture time.

With respect to your exposure choice on the merganser, I would say you did a perfect job both at capture time and during post-processing. In fact, I would argue that you did apply a bit of ETTR in that you exposed the whites a bit brighter than you actually wanted them in the final image without losing any detail. This gave you a bit less noise in the blacks so you could lighten them more in post. Generally speaking, when you have very bright whites, ETTR will provide little, if any advantage compared to exposing correctly and you also run the risk of overexposing the whites and ruining the image. But when you have a special case like this one or with a loon where you have very bright whites as well as very deep blacks, a bit of ETTR is key to bringing out the detail in the blacks with less noise. However, you need to be very careful as you are working at the bleeding edge of exposure and still need to be able to recover the white detail in post. On the other hand, if you expose to get more detail in the whites at capture time (i.e. underexpose a bit), it will likely be very difficult to bring out the black detail without significant noise issues.

Hope that all makes sense … looking forward to hearing what others have to say.

[Vivaldo Damilano]

Excellent example Doug. I would also expose for the whites like you have explained. The latest camera's are so much better in the shadow area which make it a lot easier to recover shadow detail. Great image, love the colours and sharp detail.

[Michael Gerald-Yamasaki]

Doug,

Greetings. First the reasoning: Color, encoded in rgb values decreases in resolution, as you approach 255, 255, 255 (8-bit values) and on the other side 0, 0, 0. If the intent is to pull down whites after ETTR, one does so with the limited color resolution of near whites resulting in odd coloring. Pulling up blacks on the other hand also emphasizes the limited color resolution of near blacks and results in an increase in the appearance of color noise. This latter impact can be seen in the post processing in frame 2 (color noise in the pulled up blacks). If you were to take the image and pull the whites down further, toward the back of the head you would begin to see a greenish cast as an artifact of the limited color resolution in the whites.

To summarize, adjusting near white and near black tonal values, due to limited color resolution, results in color artifacts. ETTR trades the impact in whites (less color resolution) for impact on blacks (less noise).

So one picks their poison. Because I think seeing detail in whites is always more difficult, I would lower the exposure in this shot, a small amount as you're not completely ettr. A half a stop? On the theory that I would much rather pull up whites (from higher color resolution) than pull them down (from lower color resolution). The blacks ... Color noise reduction for the darks is quite good these days ;-) and the sensors are also pretty good in the darks for noise. Your shot for instance, even pulling up the blacks shows no evidence of luminance noise. The thinking would change some if the main subject were of primarily midtones.

A color resolution chart appears in frame 8 of this discussion: http://www.birdphotographers.net/forums/showthread.php/104075-ETTR-amp-Perception-of-Whites
Frame 1 might also interest regarding perception at near whites. I have to say though that the discussion in this thread wonders hither and yon and may not be of much interest (ymmv).

Cheers,

-Michael-

[Doug Brown]

This is a very good discussion. I hope more of you will weigh in with your own opinions.

[dankearl]

I shoot a lot of hoodies and buffleheads which are even harder to get right.
I can't ETTR and get decent results at all.
I don't know about Canon, but with Nikon, I can not have any blinkies at all and be satisfied, so I shoot
2/3 to a full stop under for Buffleheads and Hoodies.
I can bring up almost anything with the D800, but cannot fix blown whites (can anyone?).
I think ETTR works for Landscape, but birds w/ white highlights, not at all.
I think with todays sensors it is not necessary anyway. I shoot a lot of waterfalls with long exposures where you have to shoot a stop under
at least or they are ruined. The OOC images are not pretty, but D800 files are easy to fix.
I shoot water and ducks with nothing actually touching the left, but it is mostly all right at the edge of the left side.

[Bob Serling]

I love this shot, especially the beak where it can be very difficult not to get a lot of reflection.

The one question I'm curious about is why you used the 400mm f/5.6? I love that lens, but the minimum focus distance is quite long. This shot appears like you were pretty close and I would have thought you'd use the 600mm.

[gail bisson]

Firstly, I love the portrait. Beautifully exposed and superb PP work. Tack sharp as well.
I am going to take the lazy woman's way out and say that I pretty much agree with Mike M and always try to expose in the way he describes.
A word of caution though. If one exposes for the whites by checking the JPEG on the LCD display for blinkies then make sure you have the contrast set at 0 or -1. On my trip to Alaska, I set my JPEG picture style to standard (which is fine)but somehow custom set the contrast to +4. This meant that I was getting blinkies on my whites when in reality the whites in my RAW were fine and consequently I underexposed most of my images for 2 full days. Thankfully, the 1DX and PP saved my butt on most shots but some I lost.
Gail

[Miguel Palaviccini]

Nice discussion here. I've been shooting ETTR since I read about it on the forums here (so about a year now). I think one of the main points that Doug is trying to point out (correct me if I'm wrong), is that the ETTR exposure comes at a cost - and that cost was shutter speed in my case. I was aiming for a better exposure (or I should say - a file with more information) but was losing shutter speed, and hence wasn't always obtaining crisp files. There is a balance for sure. For the time being, I may exposed about 2/3 under what ETTR would tell me, giving me more SS. I should be able to correct for the "underexposure" (in quotes because the exposure may still be correct, just not as far to the right as ETTR says to shoot) in PP.

[Loi Nguyen]

Great discussion here. For this particular image, I think your blacks look better than your whites. In the OP, you had superb details on the blacks and good details on the white at the bottom, but you don't have a lot of details of the whites on top. So, IMHO, you sacrificed the whites somewhat to expose for the blacks here. With 20-20 hindsight, it would have been great had you bracketed the exposure and then processed them to see which exposure would give you the most pleasing image.

For me, I tend to over-expose the whites more than under-expose the blacks, so I would get worried when the histogram of my images bumping against the far right edge and I would back off by 1/3 if I begin to see blinkies.

Loi

[Doug Brown]

Hi Bob. The 400 is a great lens at the Albuquerque Zoo when you're photographing ducks. I actually had my 600 with me but used the 400 for most of the morning. As you point out the MFD isn't that great, but it fits between fence posts at the pond and you can get full frame shots of ducks with it. The image I posted above represents about 80% of horizontal full frame.

[David Salem]

Hey Doug, Hope all is well.
Great topic and something I have been playing with for years. I shoot allot of contrasty subjects like Falcons, Terns, Skimmers, Ducks, Stilts, Shrikes and just last week the target species in Wyoming was Magpies.
I remember You , Jim and I standing at BC and looking at our camera settings one evening and having a similar talk about this subject. We looked at each others settings and I showed you how I liked to expose for Terns, a hare hot with just the smallest area blinking on the highlight alert.
Even though many people will tell you to expose only for the whites and then bring back the blacks in PP, allot of times, it has more of a price to pay than just taming down a small area of whites. Noise is added and detail and color can be lost.
I agree with Mike on this and its how I learned to expose by trial and error over the years.
P.S. The Hoddie looks great and thats how you want a shot to finish up, however you get there. TFS

[Karl Egressy]

Hi Doug,
The Hoodi looks great and as David so eloquently put it, what matters the most is that you get it right at the end.
I won't even mention how I shoot as I might expose myself to heavy critics.
All I know is that ninety percent of my shooting has the right exposure and I learned it by listening to experienced people like you and others as well as trials and errors on my own.

[Daniel Cadieux]

I'm a big fan of ETTR too, and yes, I can bring back details on whites that were beginning to blink on the LCD (with the 7D). My strategy is most often to expose for the whites as close as possible to the right edge of the histogram, or at least well into the fifth box, and let the shadows fall wherever they will. In a situation like this, if flash is allowed or feasible, you can also reduce your ETTR'd exposure by about a third of a stop or two and use fill-flash (NOT full flash)...looks good on the blacks IMO. Loss of SS is a non-issue in sunny conditions as it is most often plenty high anyhow, but yes this method can be a restriction in low-light or overcast days when photographing moving subject, in which case a compromise can be had by not being so aggressive on pushing the exposure to the right....but with a species like the Hooded Merganser it is easier to expose in overcast light anyhow. Awesome discussion.

BTW, very nice image!

[shane shacaluga]

Great information here. Thanks a lot for sharing

Can anyone tell me the maximum RGB numbers i should have on my whites during ACR conversion to ensure they are not blown?

Above what RGB number would blinkies on the whites appear?

[Mike Milicia]

Can anyone tell me the maximum RGB numbers i should have on my whites during ACR conversion to ensure they are not blown?

As long as no channel is showing a value of 255 on any pixel, nothing is blown out. The detail is there but the white may be so bright so as to appear hot or may even appear to be blown out and lacking detail. But as long as no channel has reached its maximum value of 255, the detail is recoverable by any number of means including reducing Exposure, turning down the Whites slider, or using a tone curve. You can also selectively recover white detail by using the Adjustment Brush and the Exposure slider.



Above what RGB number would blinkies on the whites appear?

This is a bit more complicated due to the way the camera determines whether or not to have a pixel blink. By default, most cameras including Canon and Nikon DSLRs compute the "luminosity" of each pixel to determine whether or not it should blink. The luminosity is calculated according to a formula which is designed to give the relative intensity of a pixel as perceived by human vision. Since human vision is much more sensitive to green than red or blue, the luminosity formula for a given RGB value is :

(0.3 * R) + (0.59 * G) + (0.11 * B)

A given pixel will blink if and only if its luminosity value exceeds a threshold established by the particular camera body. I have not tested recent models but past experimentation showed that the threshold for the Canon 1D Mk III was 249 and the threshold for the Canon G10 was 251. The problem with this approach is that you can easily blow one channel and still not get blinkies. For example, an RGB value of 255,180,203 generates a luminosity value of 205 which is well below the threshold for blinkies. For Canon cameras, the story ends there. With Canon, blinkies are good at warning you when all 3 channels are blown, i.e. blown whites, but they are generally useless for determining if a single channel has been overexposed. With nature photography, this can especially be a problem with intense reds such as fall color or a cardinal. You can easily toast the red channel without ever seeing a blinkie. With Canon, the presence of blinkies indicate a potential problem but the absence of blinkies does not tell you anything unless you have whites in the scene where all of red, green, and blue move in unison. The only way to reliably monitor whether you are blowing any one of the 3 channels on a Canon body is to examine each of the individual red, green, and blue histograms. The single color or "Brightness" histogram also uses luminosity values so it has all of the same issues as blinkies do.

Some (maybe all?) Nikon bodies provide a way around this issue with their "Highlights" display mode where you can request that the camera show you blinkies on an individual channel basis rather than using luminance. The default view is still "RGB" which uses luminance but you can optionally choose to view blinkies for an individual channel. For example, you can choose to view blinkies for the red channel and you will get a blinkie for any pixel whose red value is 255.

You can easily observe all of this by photographing a bright red object with no whites in the scene and watching the red histogram relative to the brightness histogram and blinkies.

[shane shacaluga]

Thanks a million! Much appreciated ;)

[arash_hazeghi]

It depends on camera.

With a camera like D800 the shadow noise is almost zero so you can pull up the dark with no penalty to the limit of hitting the noise floor (sever under exposure). there is no advantage in pushing whites and risk blowing them.

A camera like 7D suffers badly from shadow noise and pattern noise, with that body I'd even scarifies a bit in highlights to get the signal higher in the shadows, push yo the right as much as you can.


The 1DX doesn't have much shadow or pattern noise (if you use the right conversion software) so I wouldn't push the exposure and risk loosing details in the whites


hope this helps


Doug's image has great detail in whites and darks

[Rachel Hollander]

Just wanted to say this is a great discussion and perhaps should be copied to the tutorial and educational forum.

Rachel

[Dennis Luz]

I don't understand he concept of ETTR unless it is explained in simpler terms. Is the concept analogous to the old film era rule of thumb of "exposing for the shows and developing for the highlights?"

[arash_hazeghi]

I don't understand he concept of ETTR unless it is explained in simpler terms. Is the concept analogous to the old film era rule of thumb of "exposing for the shows and developing for the highlights?"

Hi Dennis,

yes, ETTR is basically pushing the exposure as high as possible without blowing the highlights. The idea is to let the sensor collect as much light as possible so the signal to noise ratio improves in the dark/shadow parts without the highlight pixels saturating and losing detail. It was a necessity in the early digital days when CCD sensors were used which suffered too much from pattern noise and shadow noise. Today's cameras have improved a lot, especially with modern high-end cameras such as D800, D4, 1DX you can pull up the shadows in post processing (within reasonable limits of course) without detrimental effects so it's not worth risking to blow the highlights. It still does have advantages with small sensor / lower end cameras.

[Dennis Luz]

Thanks Arash. BTW my typo on the word "show" should have been "shadow," but i am sure most readers figured that one out.
Arash I still remember your fine shot of the Hoodie in Stow Lake, Golden Gate park, San Francisco a couple of years back. It prompted me to go there and I saw a pair of Hoodies in Lloyd Lake, also in the park last year.

[Doug Brown]

Thanks Arash. I've been traveling all week and haven't had time to contribute to the thread like I should have.

[Doug Brown]

On the other hand, if the brightest thing in the scene is a mid-tone, ETTR can easily be applied and will greatly reduce noise in the image. You can overexpose such a scene by as much as 2 stops and not lose any information since the brightest thing in the scene was a midtone. If you do this and then bring the exposure back down 2 stops in RAW conversion, you will end up with the same image but with far less noise than if you had exposed it "correctly" at capture time.

Thanks for your excellent reply Mike! With regard to your statement that I quoted above, I have the following thoughts. I don't doubt that ETTR can significantly reduce noise. But I would rather apply those 2 stops of exposure headroom to more DOF, faster shutter speed, or lower ISO. Noise is easily managed in post processing (with rare exceptions), but there's no substitute for enough shutter speed to freeze the action or enough DOF to get the eyes and beak in focus.