Salvage job?

[Ronald Zigler]

Canon 60D, 300mm f4L + 1.4x, 1/1000, f6.3, ISO 200. I have begun to wonder about the lengths one may go to salvage an image. I am not sure how well I have done this or whether this image's limitations are too formidable. I removed some distracting branches that were out of focus.Yet, the image is not as sharp as I would like. There is perhaps too much noise since I had to lighten the shadows covering the birds belly. On the other hand, I don't know when I will ever come close to taking a picture like this again. This eagle was very close. The image is not cropped very much at all.

[Jeff Cashdollar]

Ronald,

One of the advantages of many members is we all have slightly different opinions, I am not big on saving images. I would rather spend time trying to get a better one-there is a balance. That being said, this image lacks light on the main body as you know and will not be sharp or detailed. It is a picture I would not spend time with, but each photographer must make that decision and the light on the tail is actually very nice and original. Thanks for sharing, was this a PA eagle.

[Ronald Zigler]

Jeff, I tend to agree with you. I'd like to learn how to work with RAW files and make the most of some compelling photos that otherwise fall short of the ideal; but another part of me feels I need to learn how to get it right when I am taking the picture (which I am still learning how to do). The picture of the eagle was taken at the Conowingo Dam in Maryland, just across the PA border, last month on December 22 (about 2 hours from where I live in PA). I had a pretty good day viewing and photographing the eagles. However, I was told that the best time to go is during the week of Thanksgiving in November. That's when I plan to go back.

Ronald,

One of the advantages of many members is we all have slightly different opinions, I am not big on saving images. I would rather spend time trying to get a better one-there is a balance. That being said, this image lacks light on the main body as you know and will not be sharp or detailed. It is a picture I would not spend time with, but each photographer must make that decision and the light on the tail is actually very nice and original. Thanks for sharing, was this a PA eagle.

[John Chardine]

Hi Ronald- Thinking of in-camera first, I would have used 400 or 800 ISO for a bird in flight. That way you would have been able to use a reasonably high shutter speed and expose the eagle properly. Against the blue sky, with no indicated exposure compensation, the eagle would have been underexposed, which resulted in the noise seen here.

You may be able to run some noise reduction on the image but you run the risk of losing any feather detail you have. The slight motion blur you seem to have is not easy to remove. With any flying bird I would try to obtain images of them flying more or less towards you (at least with a vector towards you), rather than overhead, flying away. With longer teles, you can compensate for their flight height by shooting them at a distance- that way the actual camera angle is pretty shallow. Snow on the ground is particularly effective with eagles because it reflects light back up underneath the subject.

[Jeff Cashdollar]

Good points by John,
Your camera is exposing for the brighter sky, thus underexposing your subject. You need to dial in some exposure compensation to correct for this, unfortunately this will result in overexposed skies! Alternatively if the subject is close enough you can use flash. Maybe a +1.5 to start with.

[Roger Clark]

John,

I'm wondering about your comment about proper exposure. If the exposure was not fast enough, switching to a higher ISO and a faster shutter speed would not have helped the noise situation. Ron was within 1/3 stop of wide open. Exposure (the amount of light recorded) in digital cameras on this kind of subject is dependent on f/ratio and exposure time and is independent of ISO as iSO does not actually change how many photons get captured (sensitivity does not change ISO). Thus, for the given exposure, Ron probably got the highest signal-to-noise ratio he could have gotten except by opening up to f/5.6, which would have improved the signal-to-noise ratio by only about 11%. A higher ISO would have helped the deepest shadows where read noise can be a factor. So perhaps 1/1000 sec. at f/5.6 and ISO 400 would have been better assuming nothing got clipped.

Edit: Some of the noise could be due to the image being jpeg and not a raw capture.

Regarding salvage or not, for me it would depend on how many other images of eagles I had. If it was the only/first one, I would put some effort into it. Either way, a learning experience.

Roger

[John Chardine]

Thanks Roger. My logic was flawed. I was under the mistaken impression that if you expose something properly that you will see less noise than if you brighten an image with the Exposure slider or Levels/Curves. What you are saying is that if you keep the exposure techs the same but just change ISO that you collect the same number of photons and the signal to noise ratio remains the same. Brightening an underexposed image simply reveals the noise that would have been there anyway, had you exposed properly by changing just ISO.

Thinking about why I thought the way I did- in the back of my mind is the mantra "expose to the right (ETTR)" because more information is contained in the right hand side of the histogram than the left. Is this simply the result of the number of photons collected or is there something in the way the information is stored (ref. http://www.luminous-landscape.com/tu...-right.shtml)? So, if you expose to the right simply by changing ISO (keep amount of light collected the same) will you gain any benefit re. S/N ratio?

Just read this very interesting link:
http://chromasoft.blogspot.com/2009/...ain-wrong.html

which relates directly to my last question above.

Update- I've done more reading on this issue on the web and it is indeed a gigantic "can or worms" and throws the efficacy of ETTR into question.

And here's a link suggesting that my flawed logic may not be so flawed!
http://www.luminous-landscape.com/fo...0277#msg460277

Now I'm confused!

[Roger Clark]

Thanks Roger. My logic was flawed. I was under the mistaken impression that if you expose something properly that you will see less noise than if you brighten an image with the Exposure slider or Levels/Curves. What you are saying is that if you keep the exposure techs the same but just change ISO that you collect the same number of photons and the signal to noise ratio remains the same. Brightening an underexposed image simply reveals the noise that would have been there anyway, had you exposed properly by changing just ISO.

Thinking about why I thought the way I did- in the back of my mind is the mantra "expose to the right (ETTR)" because more information is contained in the right hand side of the histogram than the left. Is this simply the result of the number of photons collected or is there something in the way the information is stored (ref. http://www.luminous-landscape.com/tutorials/expose-right.shtml)? So, if you expose to the right simply by changing ISO (keep amount of light collected the same) will you gain any benefit re. S/N ratio?

Just read this very interesting link:
http://chromasoft.blogspot.com/2009/...ain-wrong.html

which relates directly to my last question above.

Update- I've done more reading on this issue on the web and it is indeed a gigantic "can or worms" and throws the efficacy of ETTR into question.

And here's a link suggesting that my flawed logic may not be so flawed!
http://www.luminous-landscape.com/fo...0277#msg460277

Now I'm confused!

Hi John,

Lot's to discuss. The bottom line is all of the above have components that are wrong, which, like much on the internet adds to a lot of confusion.

The basic premise of expose to the right is correct, and dates back decades and did not originate with Reichman, no matter what may be claimed on the net. Reichman just popularized it for photographers. Basically for each stop in longer exposure, one doubles the collected photons, increasing signal-to-noise ratio (S/N) by square root 2. As pointed out on the second link, root 2 gains are small but can be seen in images. The idea of more A/D levels is totally bogus; it is all about photons. The second link on "ETTR is wrong" has many conceptual errors and mixes images with in-camera noise reduction to confuse the issue further, plus using jpegs with its compression noise to reduce apparent effects more.

So more exposure = higher S/N and better image regardless of ISO as long as important parts of the image do not saturate (caveat below).

Changing ISO does not change sensitivity; it only instructs the camera to expose shorter and apply more post sensor gain as one increases ISO.

But increasing ISO is a major benefit aluded to in the third link. Digital camera A/D converters have a noise floor about 11 to 12 stops below max. So at low ISO (and that is camera dependent), the sensor has a higher dynamic range than the A/D can handle. So at low ISO the noise in the shadows can be dominated by A/D noise. At higher ISOs (e.g. 400 to 800 on 1DIV, 7D seried, higher on older generation DSLR cameras) the post sensor gain boosts signal above the A/D noise floor. So the effects shown in the 3rd link with higher versus lower ISO only is valid at low ISO. The system performance is a balance between signal (light) and noise sources: photon noise (which is the squarer root of the signal level), A/D noise, and sensor read noise.

So the bottom line is that for a given exposure (e.g. say one needs 1/2000 second for a BIF), use the highest ISO that does not saturate the signal at that exposure (expose to the right in whatever ISO is used). In cases where exposure time is not a factor (e.g. landscape on a tripod) usually lowest ISO is best to give the longest exposure.

A caveat is high dynamic range in a single exposure might not be as good at the lowest ISO. For example, I do a lot of night scenes with my 5D2. Like 30 second exposures at ISO 100. Then in post processing I would boost the darker areas. What I found is that there is low level banding (fixed pattern) noise like that seen in your third link. That is noise from the A/D system and will not show at higher ISO. So I do night scenes at ISO 400 and higher (even 3200) on the 5D2. The 1DIV, being a newer generation is much better and can work better at lower ISOs and still boost the shadows (7D too). I had hoped Canon would address this problem (it has been an issue since the 2002 era with the D60). They are doing better, but still have a ways to go. The next major breakthrough will be 16-bit A/Ds and allow "ISO" selection in post processing with a very clean low end.

Exposure compensation: better to change ISO (especially when in manual) than to underexpose and boost in post processing. But only if one doesn't clip highlights.

Ideally, what would I have done with the image in this thread (if I could have reacted fast enough)? Exposure times needed to be faster, so boost exposure to 1/2000 second, then f/5.6 (more light) and ISO 800 (at least 400). The overall image would be noisier (less light recorded) but shadows might be better due to A/D noise would be less of a factor.

Roger

[Roger Clark]

Ron sent me his original image (I offered to look at it if he was willing). The original capture was jpeg and in the dark areas under the wing, the levels were only 1, 2 or 3 out of 255, which with added noise is not enough quality data to show much. A raw capture would have provided more levels so helped with the tonality but the noise would still be high (though should be better). I would say the image was underexposed by at least 2 stops. So increasing ISO to 800, opening to f/5.6, recording raw and keeping the exposure at the 1/1000 second would have given a better chance at a image that might work. But the light was pretty difficult. Better yet in this case (due to the backlit subject) would have been flash (e.g. with better beamer). A faster shutter speed would have only hurt the situation (regardless of ISO), which makes the subject a tough catch. There was motion blur saying faster shutter, but the light simply was not there.

One other thing in the image: chromatic aberration. Ron was the 1.4x a canon TC or other brand? The chromatic aberration can be easily corrected in raw converters like CS5, and also with a jpeg with some software (never tried a jpeg but in principle it is straightforward and could be done in photoshop by resampling each channel with some trials).

So Ron, I think this one is a chalk it up to a learning experience. Skip salvage, and make it an excuse to go back and try again (that's what I tell myself when I mess up--another excuse to get outside). After a few more eagle images, you'll forget about this one.

Roger

[John Chardine]

Thanks Roger. That was a great summary and I know more about this subject today than I did yesterday!

[Ronald Zigler]

The extender was a Canon 1.4x II. All equipment is new, but recently received (and needed) some adjustments performed by a Canon service center. Focusing had been inconsistent.