The frustrations and joys of a 5DmkIII in the same picture

[krisbell]

I'm sure this is new to absolutely no-one but I recently came across an example from my own library that illustrates how Canon can be so good and so bad...within the same picture.

Attached are crops from a picture taken at ISO 100, f9, 1/200 with a 100mm macro. The good part is the incredible detail and sharpness on the in-focus, well-lit areas. The less good part is the obvious noise and banding at base ISO in the darker areas. Pushing the EV or shadows by any amount whatsoever results in truly horrific noise. The crop of the dark area here is with absolutely no adjustments applied in RAW. I have included the final photo for reasons of vanity.

Despite this example, I still love my Canon and it remains my go-to system, even though I also own and love my Sony a7r.

Arabian horned viper (Cerastes gasperettii) by Kristian Bell, on Flickr

 

[scyrene]

I guess we must have different definitions of 'horrific'.

 

[krisbell]

Thanks for your comment scyrene, though my mention of 'horrific' was in relation to pushing the shadows to any degree in post.

IMHO, for what was up until very recently Canon's highest resolution pro camera, at base ISO I find the level of noise and banding to be very poor. I wouldnt be surprised if this image (content aside) would get turned down from high quality stock sites due to excessive noise/artefacts.

 

[stefsan]

I actually struggle to see banding noise in your clip from the darker background (might be just my poor monitor or my higher tolerance for noise?). I actually would gladly trade the noise in my cameras (7D and 7DII) for this one – although the 7DII is MUCH better than the original one…
At least, in shots like this it is quite easy to get rid of the noise in post

 

[stefsan]

Oh, and by the way – that is a very cool picture of an impressive animal!

 

[krisbell]

Hi Stefsan and thanks for your comments. I am sure it is better than a 7D or 7DII but that doesnt make it acceptable!

I have to say the noise doesnt look quite as bad on this crop as it did on my screen in photoshop for whatever reason?! But it was damn near impossible to get rid of in post. Every attempt (Noiseware, Dfine, Photoshop reduce noise) created horrible posterization - if you have a decent technique to remove this I'd love to hear it!

 

[GMCPhotographics]

Firstly, the tungsten light is quite poor...and offers only a limited band of the full color spectrum. So it's not a surprise that there is a little sensor noise. The sensor is working hard to produce a well exposed image to start with.
It kind of reminds me of the amount of times I see guys calibrating their AF at night, at home under tungsten light and then wondering why their AF is off a few days later (at the weekend) when they are shooting during the day under natural light. Light sources are not equal and neither it's their color band.

Try this same experiment under blue light and the iso noise would be a lot worse.

 

[krisbell]

Many thanks for your reply GMCPhotographics - can you clarify a little more about what you mean when you say the sensor is working hard to produce a well exposed image?

I think I understand what you are saying regarding the limited spectrum tungsten lights put out, but does this directly relate to noise? In terms of overall light in the picture, there was plenty enough to create a slight overexposure at base ISO. Will tungsten light sources always produce this sort of outcome?

 

[cervantes]

I absolutely love my 5D3 in about every regard, but I was shocked when I saw these comparisons with the Nikon D750:

http://petapixel.com/2014/10/14/nikon-d750-review-nikon-youve-created-monster/

Shadow noise wise this is, in my opinion, not a comparison between two DSLRs but between a P&S and a DSLR!!

I will not buy another Canon camera until Canon is on par with Sony, sensor-wise (or if they use Sony sensos for that matter).

 

[K]

Here's my $0.02

I think the image is excellent with good quality. The 5D3 is a do-everything-well camera that isn't perfect in any one area. It is impressive to put out this kind of quality for what it is.

You yourself, and perhaps some pixel peeping fanatics on the web might take issue with the noise, but that's about it. It isn't perfect at ISO 100 at 1:1, but it is far from "horrific" ... It all depends what you're going to do with this image. I think it is nitpicking. The overall image is excellent and I'd wager most people would never notice, nor care even if they did. Not sure what sizes of print your offering or how it is intended to be used - but it shouldn't be an issue unless its going to become a wall sized print. And even then, viewing distance becomes proportionate to the image. It is all relative.

As others have stated, improved lighting goes a long way to making sure the sensor is properly fed. And a well fed sensor very much improves IQ.

Overall though,

If standards are that exacting that you consider this image to contain "horrific noise" - then the only thing that will truly satisfy will be a move to medium format with top of the line lighting to go with it. The 5DS or D810 will definitely improve, but it will be incremental and not a game-changer.

 

[K]

I absolutely love my 5D3 in about every regard, but I was shocked when I saw these comparisons with the Nikon D750:

http://petapixel.com/2014/10/14/nikon-d750-review-nikon-youve-created-monster/

Shadow noise wise this is, in my opinion, not a comparison between two DSLRs but between a P&S and a DSLR!!

I will not buy another Canon camera until Canon is on par with Sony, sensor-wise (or if they use Sony sensos for that matter).

Wonderful, another one of these "5 stop push" comparisons -

While impressive of the sensor - the final image still looks like garbage.

A novelty at best.

I still await some Nikonian to prove how this amazing 5-stop shadow lifting ability translates into real advantages within high quality, real photographs. I have yet to see this. This of course, is because whether it is Nikon, Canon or Sony - one can only really lift shadows about 1/2 stop before IQ suffers. The mighty Exmor does not overule the fundamentals of photography which dictates the exposure must be perfect to maximize IQ.


There are 2 issues with shadow lifting.

1. Lifting them more than a stop creates unrealistic, bizarre looking images. The real world has shadows and dark areas. It is actually a good thing! Gives images depth. The majority of pros who use Canon know this, and don't create flat looking, unrealistic images like all the Nikon amateurs out there with all their shadow lifting lightroom antics. Pros are less obsessed with sensors and DR, and completely and totally obsessed with lighting which is what creates the image.

2. Anything more than a stop degrades the image quality, even for the mighty Nikon Sony Exmor sensor. If that much detail is wanted in the shadows, then the exposure should be adjusted for that. Because it isn't adjusted for the shadows - the exposure is wrong for them, and even a 1 stop lift degrades the IQ. In my opinion, this is HDR territory.

If the 5D4 has 18 stops of DR, it would not change my opinion. The exposure must be correct to optimize IQ. But everyone has different standards for what makes a great image.

It is obvious the Exmor worshipers haven't spent hours pixel peeping high resolution digital medium format images produced by professionals with high end lighting. If they did, they'd quickly understand how the high DR Exmor is nice, just not as practical as they think it is.

 

[krisbell]

Many thanks for the link Cervantes - it is a similar story to what I have seen personally.

K - thanks for your thoughts. Once again I re-iterate that I do not consider the noise "horrific" as displayed, merely disappointing. And I will re-iterate that I find the overall quality of detail and colour etc to be excellent. HOWEVER, when pushed in post to any extent the noise and banding quickly becomes horrific. I agree any sensor has its limits, and each persons mileage will vary, but at ISO 100 this image has no leeway whatsoever. For me this image is a classic example where a 1 stop push by a Sony sensor is extremely valuable. I dont want to raise things dramatically but just lift the darks at the top of the image just a little. It wont save a terrible picture but it can help elevate an image to the next little level (i.e. make an average image good, a good image great, a great image a masterpiece etc).

 

[GmwDarkroom]

I think I understand what you are saying regarding the limited spectrum tungsten lights put out, but does this directly relate to noise? In terms of overall light in the picture, there was plenty enough to create a slight overexposure at base ISO. Will tungsten light sources always produce this sort of outcome?

Tungsten light is a very warm, red cast and can have 3-4x as much red wavelength intensity as blue. Any time your lighting is more intense in one area of the spectrum (Tungsten, Mercury Vapor, etc), your camera sensor which uses different layers/sites to capture different portions of the spectrum will have some layers/sites that are receiving good light intensity with low noise and others that are receiving low light intensity and thus producing more noise. As you get into the shadows of those photos, even less light of the already starved areas of the spectrum exists and you'll be pulling in even more noise from those areas of the sensor. In well lit areas, all sites get sufficient light to produce negligible noise so the result can be shadows that look that much worse.

If your lighting was an even spectrum or your camera sensor captured all wavelengths at all capture sites, you wouldn't have this problem. While you might get a better shot out of a camera with lower noise, you'll still have more noise than if you had an even spectrum of light. There's only so much the camera can do to correct this.

 

[krisbell]

Thanks Gmw - this makes perfect sense and is something that until today had never crossed my mind. I think it explains much of what I am seeing. Is there anything I can do in order to 'fix' this going forward - would it help if I used blue gels on the flash or added some blue info/noise in post for example?

 

[Andyx01]

I laughed.

I cryed.

It was the best of times.

It was the worst of times.

But no seriously; I would rather have 'noise' then an overly proccessed smear job of gradient areas that many other bodies do to get better ratings from people that think they know better.

 

[scyrene]

Thanks for your comment scyrene, though my mention of 'horrific' was in relation to pushing the shadows to any degree in post.

IMHO, for what was up until very recently Canon's highest resolution pro camera, at base ISO I find the level of noise and banding to be very poor. I wouldnt be surprised if this image (content aside) would get turned down from high quality stock sites due to excessive noise/artefacts.

The question is, why are you pushing the shadows? The shot is perfectly exposed as far as I can see.

But I have had to deal with noisy backgrounds with smooth gradients a fair bit, as many of my bird shots are at high ISO. My technique (using Lightroom) - if you want to eliminate it as much as possible - is to use a brush tool to paint the out of focus areas - use denoise *and* lower the clarity/sharpness, so it smooths out any posterisation. It takes longer, but I'd only be doing this for shots I wanted to print or show off large. Reducing to web viewing size reduces the noise massively, of course.

 

[scyrene]

I absolutely love my 5D3 in about every regard, but I was shocked when I saw these comparisons with the Nikon D750:

http://petapixel.com/2014/10/14/nikon-d750-review-nikon-youve-created-monster/

Shadow noise wise this is, in my opinion, not a comparison between two DSLRs but between a P&S and a DSLR!!

I will not buy another Canon camera until Canon is on par with Sony, sensor-wise (or if they use Sony sensos for that matter).

Wonderful, another one of these "5 stop push" comparisons -

While impressive of the sensor - the final image still looks like garbage.

A novelty at best.

I still await some Nikonian to prove how this amazing 5-stop shadow lifting ability translates into real advantages within high quality, real photographs. I have yet to see this. This of course, is because whether it is Nikon, Canon or Sony - one can only really lift shadows about 1/2 stop before IQ suffers. The mighty Exmor does not overule the fundamentals of photography which dictates the exposure must be perfect to maximize IQ.


There are 2 issues with shadow lifting.

1. Lifting them more than a stop creates unrealistic, bizarre looking images. The real world has shadows and dark areas. It is actually a good thing! Gives images depth. The majority of pros who use Canon know this, and don't create flat looking, unrealistic images like all the Nikon amateurs out there with all their shadow lifting lightroom antics. Pros are less obsessed with sensors and DR, and completely and totally obsessed with lighting which is what creates the image.

2. Anything more than a stop degrades the image quality, even for the mighty Nikon Sony Exmor sensor. If that much detail is wanted in the shadows, then the exposure should be adjusted for that. Because it isn't adjusted for the shadows - the exposure is wrong for them, and even a 1 stop lift degrades the IQ. In my opinion, this is HDR territory.

If the 5D4 has 18 stops of DR, it would not change my opinion. The exposure must be correct to optimize IQ. But everyone has different standards for what makes a great image.

It is obvious the Exmor worshipers haven't spent hours pixel peeping high resolution digital medium format images produced by professionals with high end lighting. If they did, they'd quickly understand how the high DR Exmor is nice, just not as practical as they think it is.

Quite. If you want to reduce shadow noise in shots like this, ETTR and reduce the overall exposure (fiddling with contrast as required, etc). Underexposing at any ISO is going to increase noise, and is just bad technique (though obviously in scenes with high contrast it is harder to avoid).

 

[scyrene]

Many thanks for the link Cervantes - it is a similar story to what I have seen personally.

K - thanks for your thoughts. Once again I re-iterate that I do not consider the noise "horrific" as displayed, merely disappointing. And I will re-iterate that I find the overall quality of detail and colour etc to be excellent. HOWEVER, when pushed in post to any extent the noise and banding quickly becomes horrific. I agree any sensor has its limits, and each persons mileage will vary, but at ISO 100 this image has no leeway whatsoever. For me this image is a classic example where a 1 stop push by a Sony sensor is extremely valuable. I dont want to raise things dramatically but just lift the darks at the top of the image just a little. It wont save a terrible picture but it can help elevate an image to the next little level (i.e. make an average image good, a good image great, a great image a masterpiece etc).

Following on from what I said above, I would recommend shooting at a higher ISO in that case. Using ISO 200 to slightly overexpose (without blowing highlights), then reducing the exposure will help reduce shadow noise.

 

[sanj]

I totally understand what you are saying. It is a problem indeed.

 

[krisbell]

But I have had to deal with noisy backgrounds with smooth gradients a fair bit, as many of my bird shots are at high ISO. My technique (using Lightroom) - if you want to eliminate it as much as possible - is to use a brush tool to paint the out of focus areas - use denoise *and* lower the clarity/sharpness, so it smooths out any posterisation. It takes longer, but I'd only be doing this for shots I wanted to print or show off large. Reducing to web viewing size reduces the noise massively, of course.

That is something I often do in my images but for images like this it actually makes posterisation worse!! And its not possible to tell from the crops I have provided but the image actually was ETTR. The final image is brought back down by almost half a stop. I think it mostly boils down to the earlier suggestion about poor spectrum of my flashes that has created the issue in this specific picture. Thanks once again for the comments.