Canon 5d Mark III Shadow recovery

[jrista]

Here is an example of a scene with extreme dynamic range that perfectly demonstrates the "window test" that Art_d mentioned a page or two back:

http://www.flickr.com/photos/26562595@N02/7043690229/#lightbox/

Nearly-blown highlights as well as areas that appear to be completely black, as the scene had around 13 stops of DR. I want to contrast this against the underexposed shed shot...which did not really have even moderately-bright highlights...I think pretty much every pixel was below a middle toned gray in the shed shot. That indicates it was not a DR-limited scene. The scene above, however, is definitely a DR-limited scene...you have every level from near total black to near pure white (and the pixels on the seat of the chair outside on the patio may indeed be clipped whites). This is exactly the kind of scene where having more DR than the 5D III offers is valuable.

The D800's additional two stops of DR allow both the highlights and the deep shadows to be recovered, and recovered completely cleanly, cleanly, devoid of any noise:

http://www.flickr.com/photos/26562595@N02/6897594964/#lightbox/

There was a long period of time where I wasn't really sure of the true value of the extra two stops of DR that the D800 offered until I saw these two shots. I'd seen Fred Miranda's examples, and they were pretty amazing, and that was the first time I'd seen a real-world example of Canon's ugly shadow banding noise. But it was this scene that really blew my mind. Seeing these really demonstrated the value of a near-noiseless sensor...at least 75% of the above scene would fall into "zones 1-3" (maybe only zones 1-2), and yet they recover not only well, but beautifully, and with very good color fidelity!

The RAW files for this scene used to be available here. The link is apparently dead now (they were posted a year ago), but it was amazing messing with these files. Pushing exposure around like it was entirely, losslessly fluid...the only thing I ever saw was a little bit of luminance grain in the deepest shadows (i.e. along the bed fringe of the near lower-left corner bed)...and that was it. It was a little difficult to cleanly correct both the highlights and the shadows as well as the person who posted the versions I linked above (they must have spent some careful time on them), but still...the most amazing example of shadow recovery I think I've seen.

When I first saw these photos was when I decided to wait before I bought any new Canon DSLRs for my landscape photography, and see if they could come out with something that would be capable of something similar. Even the amazing 1D X couldn't recover that scene that cleanly at ISO 100...not without visible banding and blotchy color noise.

 

[RLPhoto]

Just to buy a nikon art_d. I thought for a second there that this thread wouldn't degrade again into another DR thread.

 

[Sporgon]

Darn

Imagine if we had shot this on a D800 instead of out dated gear from 2005 :'(

 

[sach100]

@Sporgon Great shots on your website!

 

[Sporgon]

@sach100,

Thanks for that, much appreciated.

We try to shoot in dramatic light so do a lot of shadow / low lights recovery. As you have found, there's no problem in practical application.

 

[psolberg]

When I first saw these photos was when I decided to wait before I bought any new Canon DSLRs for my landscape photography, and see if they could come out with something that would be capable of something similar. Even the amazing 1D X couldn't recover that scene that cleanly at ISO 100...not without visible banding and blotchy color noise.

yup. pretty much why I shoot landscapes with a D800 exclusively. DR, color depth, resolution. It's a waste to hike miles to shoot with anything else, except off course MF.

 

[art_d]

Just to buy a nikon art_d. I thought for a second there that this thread wouldn't degrade again into another DR thread.

Why? There is more to a camera system than just the sensor. I have no desire to switch to a Nikon system and downgrade my lenses, work with poor live view, etc.

Examples were asked for. So examples were provided. And then those examples were attempted to be discredited and dismissed. As I've said, I too can post images all day long where DR is no problem at all, but that doesn't make the images when it is a limitation magically disappear.

I see nothing wrong with illustrating the scenarios where sensor limitations come into play which are based on my actual experience. Nor with expressing the opinion that if Canon could improve this area, it would be very useful.

 

[jrista]

Just to buy a nikon art_d. I thought for a second there that this thread wouldn't degrade again into another DR thread.

Why? There is more to a camera system than just the sensor. I have no desire to switch to a Nikon system and downgrade my lenses, work with poor live view, etc.

Examples were asked for. So examples were provided. And then those examples were attempted to be discredited and dismissed. As I've said, I too can post images all day long where DR is no problem at all, but that doesn't make the images when it is a limitation magically disappear.

I see nothing wrong with illustrating the scenarios where sensor limitations come into play which are based on my actual experience. Nor with expressing the opinion that if Canon could improve this area, it would be very useful.

This is it exactly. I'd love to have more DR with a Canon DSLR, but I'm not willing to give up all my excellent glass, my awesome ergonomics, or any of the other features that attract me to Canon in order to get some more DR at low ISO. It's a very nice thin to have when you need it, which is not always, but it is not quite compelling enough to make me want to switch on its own. I also don't think Canon is literally incapable of improving in this area, as some have stated in the past. I figure it won't be too terribly long before they start catching up in the DR area...maybe it will even be the 7D II.

 

[RLPhoto]

Why complain about DR any longer? Jeez, it like when people asked me why shoot slide film? It doesn't have enough DR is what they'd complain about but I loved my velvia 50.

 

[jrista]

Why complain about DR any longer? Jeez, it like when people asked me why shoot slide film? It doesn't have enough DR is what they'd complain about but I loved my velvia 50.

This debate was not about complaining about DR. Trumpet was making some rather bold claims, and dismissing the original point Art_d was trying to make. The debate has been about the validity of Trumpet's claims, as in many cases they are wildly off base. I'm a Canon fan, I love my gear and I wouldn't give up my glass for the world. I think there is more to IQ than DR (hell, I've argued that side of the fence far more than I've argued the other side), and for what I shoot most of the time, wildlife and birds, low ISO DR is a moot point anyway, and Canon has the high ISO, accurate AF, high frame rate end of things covered more than good enough.

That said, you cannot simply offhandedly dismiss the value of more DR at low ISO for certain types of photography as Trumpet has been trying to. They may be limited in number, but they certainly are not limited in scope. And we aren't talking about something on the order of a third of a stop, which would be largely irrelevant.

There is obviously a serious misunderstanding when one thinks that the difference between the 5D III and the D800 in terms of ratio, by convoluting the argument into a comparison of cars....one that goes 95mph and another that goes 105mph...as an example of the difference in DR between those two cameras. The fact that every "stop" is a doubling seems to have been lost. Two stops is a FOUR FOLD increase in dynamic range, and that is more like the difference between tops speeds of 95mph and 380mph...a very significant and very meaningful difference.

On top of that, the claims that the 5D III was capable of 30 stops of DR was just flat out ludicrous. Claims that the 5D III can capture 16 stops of DR is ludicrous. The notion that the "10-stop shed lift" resulted in "ok IQ" is also ludicrous...the noise in those images is atrocious, and completely unacceptable.

Sorry, but even though I usually like to defend Canon's IQ for what it is...excellent if no longer "the best of the best", I couldn't just ignore so many blatant misunderstandings and incorrect information lie unaddressed. I enjoy a good argument, and in this case, the Canon guy was wrong, and I responded.

 

[TrumpetPower!]

Two stops is a FOUR FOLD increase in dynamic range, and that is more like the difference between tops speeds of 95mph and 380mph...a very significant and very meaningful difference.

I've yet to meet a Nikon fanboi who has any idea what a logarithm is or why anybody would ever want to use one.

b&

 

[TrumpetPower!]

So, I tried to think of the silliest thing I'd ever actually contemplate doing in the real world. Here it is.

It's a single exposure developed twice: once for the Sun and sky, once for the foreground.

And, normally, when shooting something like this, I'd render the shadowed, backlit foreground as a couple stops underexposed. You know? Shadowed and backlit?

But, no, for the sake of silliness, I rendered it "properly" exposed -- as if I had an insanely massively huge array of octabanks aimed at that back wall. (Can you imagine the power you need to evenly and softly light up that much wall at f/11 from 30' away?)

I then blended the two exposures with a single giant gradient mask from top to bottom. I suppose I could have used a hard-edged mask on the contours of the wall to mimic the octabank, but this already looks unnaturally ugly enough to me.

A camera that can pull off something this silly in a single shot has all the dynamic range any competent photographer needs and then some. Whether the D800 can do even more is irrelevant, as the 5DIII already has ludicrous dynamic range.

TL/DR: If you can't get the shot with a 5DIII, it's your problem, not the camera's.

Cheers,

b&

 

[sdsr]

Here is an example of a scene with extreme dynamic range that perfectly demonstrates the "window test" that Art_d mentioned a page or two back:

http://www.flickr.com/photos/26562595@N02/7043690229/#lightbox/

Nearly-blown highlights as well as areas that appear to be completely black, as the scene had around 13 stops of DR. I want to contrast this against the underexposed shed shot...which did not really have even moderately-bright highlights...I think pretty much every pixel was below a middle toned gray in the shed shot. That indicates it was not a DR-limited scene. The scene above, however, is definitely a DR-limited scene...you have every level from near total black to near pure white (and the pixels on the seat of the chair outside on the patio may indeed be clipped whites). This is exactly the kind of scene where having more DR than the 5D III offers is valuable.

The D800's additional two stops of DR allow both the highlights and the deep shadows to be recovered, and recovered completely cleanly, cleanly, devoid of any noise:

http://www.flickr.com/photos/26562595@N02/6897594964/#lightbox/

Yes, it's clever that a camera allows you (or whoever's photo this is) to do that, but (and this is an aesthetic opinion more than anything else) I can't help noting that in this particular case it's been done rather heavy-handedly to what was evidently a botched (deliberately?) photo in the first place. Doesn't the "correction" look disconcertingly unreal? It's obvious from the still overblown highlights and other features of the exterior that it's bright and sunny outside, yet the exterior looks less bright than the interior (even though there doesn't seem to be any sunlight illuminating the room). If it really was brighter inside, would the original be so dark? (And could the horrible green and red fringing on the tree trunk not have been removed?)

 

[Sporgon]

@sdsr, you've got to the nub of the matter; to really try and prove the benefit of lifting under exposure on the exmor sensor these guys are having to go to extremes. Something that poor old TrumpetPower has been trying to point out until he's turned purple.

Most of us know that the Nikon has less noise/ banding etc when you really push it, but when it comes to producing real pictures the benefit is much less noticeable - unless you're main pleasure from digital photography is lifting under exposed data to mid tones and viewing it at 200%. :

 

[TrumpetPower!]

Here is an example of a scene with extreme dynamic range that perfectly demonstrates the "window test" that Art_d mentioned a page or two back:

http://www.flickr.com/photos/26562595@N02/7043690229/#lightbox/

Nearly-blown highlights as well as areas that appear to be completely black, as the scene had around 13 stops of DR. I want to contrast this against the underexposed shed shot...which did not really have even moderately-bright highlights...I think pretty much every pixel was below a middle toned gray in the shed shot. That indicates it was not a DR-limited scene. The scene above, however, is definitely a DR-limited scene...you have every level from near total black to near pure white (and the pixels on the seat of the chair outside on the patio may indeed be clipped whites). This is exactly the kind of scene where having more DR than the 5D III offers is valuable.

The D800's additional two stops of DR allow both the highlights and the deep shadows to be recovered, and recovered completely cleanly, cleanly, devoid of any noise:

http://www.flickr.com/photos/26562595@N02/6897594964/#lightbox/

Yes, it's clever that a camera allows you (or whoever's photo this is) to do that, but (and this is an aesthetic opinion more than anything else) I can't help noting that in this particular case it's been done rather heavy-handedly to what was evidently a botched (deliberately?) photo in the first place. Doesn't the "correction" look disconcertingly unreal? It's obvious from the still overblown highlights and other features of the exterior that it's bright and sunny outside, yet the exterior looks less bright than the interior (even though there doesn't seem to be any sunlight illuminating the room). If it really was brighter inside, would the original be so dark? (And could the horrible green and red fringing on the tree trunk not have been removed?)

It's even worse than you make it out. Much worse.

First, the non-lifted scene is exposed about right for the outdoors if that was all that was in the photograph. Flip between the two and you can see that the sky is actually darker and much muddier in the HDR rendition!

If you're going to do something like this, you want the outside to appear bright and the inside to appear not dark but dim. The sky should be a very light blue -- almost white, but shy enough from white to still have color in it.

Assuming, for whatever reason, that I couldn't light this scene properly, I'd take two exposures regardless of camera -- one for the outdoors and one for the interior. And the exposures I'd make would already be a bit overexposed for the outdoors and a bit underexposed for the interior -- basically exactly as they'd wind up in the final image. Editing would be nothing more than masking the two together.

Here's a quick-and-dirty edit. The shadows under the beds are still blocked, but I'm not going to waste more time fixing somebody else's bad post-processing.

And the 5DIII would have no trouble doing this shot, noiselessly, even with a single exposure if you were so lazy you couldn't press the shutter twice.

Cheers,

b&

 

[Cgdillan]

Darn

Imagine if we had shot this on a D800 instead of out dated gear from 2005 :'(

Love this shot!!!!!!

 

[art_d]

@sdsr, you've got to the nub of the matter; to really try and prove the benefit of lifting under exposure on the exmor sensor these guys are having to go to extremes. Something that poor old TrumpetPower has been trying to point out until he's turned purple.

Most of us know that the Nikon has less noise/ banding etc when you really push it, but when it comes to producing real pictures the benefit is much less noticeable - unless you're main pleasure from digital photography is lifting under exposed data to mid tones and viewing it at 200%. :

I don't consider this as "extreme":

Let me summarize the blog post which accompanied the example. The scene was exposed to capture the colors in the sky. Because of DR limitations, not blowing out the sky meant the the lower left corner of the exposure was essentially black. The goal was to lift the shadows there a bit and get to the dark blue color. This resulted in ugly fixed pattern noise in that area which was noticeable in an actual print. Also note we are not talking pushing blacks into midtones. As you can see those are still shadows that area.

We can go on all day long posting examples of where DR is not a limitation. That doesn't change anything in the photos where it is.

 

[Sporgon]

Darn

Imagine if we had shot this on a D800 instead of out dated gear from 2005 :'(

Love this shot!!!!!!

Many thanks cgDillan, it's one of our favourites.

 

[Sporgon]

@art_d, I agree that's not an extreme example of manipulation, and you haven't lifted it too much. The lighting however is quite extreme but the picture, to me, is demonstrating signs of under exposure in the first place.

If you've struggled with unacceptable results from lifting these low lights then you really have to review your technique and the programs you are using to post process.

Yourself and jrista are behaving like Nikon missionaries trying to spread your good word to the great unbelievers.

Unfortunately missionaries sometimes get eaten.

 

[Loswr]

Yourself and jrista are behaving like Nikon missionaries trying to spread your good word to the great unbelievers.

I should point out that some of us believe in DR (should I call it 'salvation' in this context?), but don't find the tradeoffs worthwhile.