Canon, Nikon Fashion/Portrait shootout

[teddillard]

)

[BernardLanguillier]

The piece I'm working on now is the comparison between the 5DMI, Nikon D3s and the Hasseys...  can't wait to pull up the ISO 800 samples of that.

Thoughts?

Why ISO 800?

On the Nikon side, I would think that the d3x is a bit superior to the D3s at ISO800, but that is probably still the ISO where the gap with the competition is the smallest.

ISO100 would show the D3x in its best light, and ISO3200 would show the D3s at its best.

Horses for courses.

Cheers,
Bernard

[teddillard]

Why ISO 800?

On the Nikon side, I would think that the d3x is a bit superior to the D3s at ISO800, but that is probably still the ISO where the gap with the competition is the smallest.

ISO100 would show the D3x in its best light, and ISO3200 would show the D3s at its best.

Horses for courses.

Cheers,
Bernard

ISO 800 simply because it was the best ISO for the shoot- balancing the tungsten source intensity and quality with the depth of field and shutter speed requirements.  Again, real shoot, not lab tests.

[dwdallam]

These types of reviews are fun to read, and, admittedly, informative, but I think they go wrong in one area: They don't take into consideration processed files. Who takes images directly from the camera to print, except for a journalist doing daily shots for the daily news papers?

Are Canon cameras noisier than the lowest noise Nikon? I could care less because the real question is: Can one camera match the low noise of another camera AFTER processing. Put another way, Nikon, etc., might have lower noise out of he camera, but can other cameras match that low noise after processing? (I'm just using noise as one an example.)

I've seen Canon ISO 800/1600 images as clean as any Nikon--after processing. So now the question changes: Would you rather have  camera that does the noise reduction for you, or allows you to keep that information and decide what you want to do with it in post?

I think I'm on the latter side of that question myself.

[Jonathan Ratzlaff]

A better comparison would have been to use the same pose, same lighting.  Because the model's pose is not the same for each image the lighting changes enough to make a difference.  The fact that this is not done, more or less negates the comparison.  It may have been an interesting exercise but a futile one.

[Rob C]

A better comparison would have been to use the same pose, same lighting.  Because the model's pose is not the same for each image the lighting changes enough to make a difference.  The fact that this is not done, more or less negates the comparison.  It may have been an interesting exercise but a futile one.

Also, comparing the left side of the face with the right side, which has different lighting, is pointless too. It is exactly the sort of 'test' that proves nothing beyond the futility of such experiments.

As mentioned in another thread today, you have to work with what you have and make the best of it once you truly understand it. That used to apply to film, in the old days, and was far easier and less expensive a problem to resolve than sensors... but let's not even go there!

;-)

Rob C

[teddillard]

These types of reviews are fun to read, and, admittedly, informative, but I think they go wrong in one area: They don't take into consideration processed files. Who takes images directly from the camera to print, except for a journalist doing daily shots for the daily news papers?

Are Canon cameras noisier than the lowest noise Nikon? I could care less because the real question is: Can one camera match the low noise of another camera AFTER processing. Put another way, Nikon, etc., might have lower noise out of he camera, but can other cameras match that low noise after processing? (I'm just using noise as one an example.)

I've seen Canon ISO 800/1600 images as clean as any Nikon--after processing. So now the question changes: Would you rather have  camera that does the noise reduction for you, or allows you to keep that information and decide what you want to do with it in post?

I think I'm on the latter side of that question myself.

heh.  Now THAT's a can of worms...  

I actually did some testing a while back, more personal than anything else, to see what I could do to process several cameras to "best advantage" using all the tools I had.  It's a great point, really, but one that I'd never post as a review, other than my blog- simply because the results are completely subjective- both in taste, but also in skill levels.  (Testing processing software, as a general subject, is a tough one as well.  More than a few times I've come to a conclusion about a processing system, only to be shown that I had missed some setting, feature or switch.  The only software that I'm really comfortable making hard and fast claims on is stuff I know really, really well.)

Obviously, the less work the file needs, the easier it is to get it to where you want it to be, and, in most of my experience, the better it's going to ultimately look.  Thus, the attempts to take processing more or less out of the equation.  The idea was to find out where the files go with the least control from the processing, to see how close you can get to what you want. Of course, some needs to be done- which is why we opted for the OEM processing vs. ACR vs. the in-camera processing (JPEG)- essentially trying to compare "generic" processing.

[teddillard]

A better comparison would have been to use the same pose, same lighting.  Because the model's pose is not the same for each image the lighting changes enough to make a difference.  The fact that this is not done, more or less negates the comparison.  It may have been an interesting exercise but a futile one.

Yeah, you know, I think that's a really good point, too.  I love the idea of a shootout, but I think we'll definitely do things a little differently next time- fewer cameras, more time with the model, I would have loved to do some outside shooting, too (the New England weather didn't cooperate).  

From my standpoint as a photographer, it was an amusing shift of gears.  When you're working with a model, you're trying to find the best pose, the best shot.  In this case, the first series with the first camera, I did all that, and then had to do it again- 8 times.  It was confusing for both me and the model.  It's an interesting problem- get some movement and life, as you would any shoot, but make it repeatable...  

Thanks for the feedback!

[Jeremy Payne]

Would you rather have  camera that does the noise reduction for you, or allows you to keep that information and decide what you want to do with it in post?

+1,000,000 ... this post should go to the Nikon vs. Canon Hall of Fame.

[teddillard]

+1,000,000 ... this post should go to the Nikon vs. Canon Hall of Fame.

sorry, huh?

...need.  more.  coffee.

[Jeremy Payne]

sorry, huh?

My apologies, Ted ... I was being snarky.

I didn't really want to pick a fight, but the idea that someone would choose Canon over Nikon because the noisier files "clean up good" just struck a chord ... seemed to me to be the epitome of these Canon vs. Nikon debates ...

[ziocan]

Good try.....
Unfortunately this test is a bit all over the place.

Don't try that at home....

[bjanes]

Are Canon cameras noisier than the lowest noise Nikon? I could care less because the real question is: Can one camera match the low noise of another camera AFTER processing. Put another way, Nikon, etc., might have lower noise out of he camera, but can other cameras match that low noise after processing? (I'm just using noise as one an example.)

I've seen Canon ISO 800/1600 images as clean as any Nikon--after processing. So now the question changes: Would you rather have  camera that does the noise reduction for you, or allows you to keep that information and decide what you want to do with it in post?

I think I'm on the latter side of that question myself.

The major sources of noise in most normal digital photography are shot noise and read noise (ignoring pixel response nonuniformity, PRNU). The shot noise is limited by the full well capacity of the sensor, and I think that Canon and Nikon are roughly equivalent in this area. The read noise is most apparent in the shadows and for a long time Canon was markedly superior to Nikon with respect to read noise. However, more recently, Nikon has closed the gap and the D3s (I believe) is superior to any existing Canon.

Some in camera noise reduction (such as double correlated sampling) can reduce read noise with no effect on image detail. This is a win-win situation. However, noise reduction via filtering does compromise image detail. As far as I know, neither Canon nor Nikon filter the raw image. PRNU can be reduced in post with no loss of image detail, but most NR in post processing is via filtering, and results in loss of image detail.  Filtering can be demonstrated via Fourier analysis and the Nikon raw images demonstrate no evidence of filtering on Fourier analysis.  What kind of NR are you doing with your Canon images?

[Jeremy Payne]

(ignoring pixel response nonuniformity, PRNU)

Ok ... so now we understand how the random poisson distribution of photons causes shot noise ... but I'm curious about the mechanism to which you are referring.

Photons arrive at the photo site and arouse electrons who flow into capacitor wells when voltage is applied, right?

Poisson distribution causes shot noise because the photons don't all arrive "as expected" ... there is some random fluctuations in their arrival patterns.

Are there also random fluctuations in the electron flow?  I would imagine yes.  

Is that the source of PRNU?  

If not ... what is the source of PRNU and what is the impact of fluctuations in the electron activity?

Sorry for my layman's terminology - it has been more than 20 years since I studied physics ...

[JeffKohn]

Are Canon cameras noisier than the lowest noise Nikon? I could care less because the real question is: Can one camera match the low noise of another camera AFTER processing. Put another way, Nikon, etc., might have lower noise out of he camera, but can other cameras match that low noise after processing? (I'm just using noise as one an example.)

I've seen Canon ISO 800/1600 images as clean as any Nikon--after processing. So now the question changes: Would you rather have  camera that does the noise reduction for you, or allows you to keep that information and decide what you want to do with it in post?

I think I'm on the latter side of that question myself.

You would have a valid point if Nikon used noise filtering in-camera on RAW files. But as Bill said, they don't.

So if you have to use noise filtering on the Canon file, that's not "just as good" as the Nikon file without filtering because noise filtering always removes some real detail along with the noise.

[ziocan]

You would have a valid point if Nikon used noise filtering in-camera on RAW files. But as Bill said, they don't.

So if you have to use noise filtering on the Canon file, that's not "just as good" as the Nikon file without filtering because noise filtering always removes some real detail along with the noise.

Well, if we keep the windows closed, we ear no noise.
Especially PRNU noise, we can forget about it.
What a liberation.
No matter what camera we are using.

[Sheldon N]

Your review includes the comment:

"Funny, but an often forgotten benefit of a high-resolution, smaller sensor- you get an automatically longer lens effect, shallower depth of field because of that"

This is somewhat misleading... Yes, the smaller sensor has less DOF if you just change camera bodies and take a shot with a tighter composition than the full frame shot. But in practice it works in reverse - you frame the shot you want with the lens that gives you the working distance you want and the smaller sensor camera ends up having more DOF. We don't automatically shoot tight crops with a small sensor cameras just because it makes the lenses "longer".

[Brent McCombs]

Your review includes the comment:

"Funny, but an often forgotten benefit of a high-resolution, smaller sensor- you get an automatically longer lens effect, shallower depth of field because of that"

This is somewhat misleading... Yes, the smaller sensor has less DOF if you just change camera bodies and take a shot with a tighter composition than the full frame shot. But in practice it works in reverse - you frame the shot you want with the lens that gives you the working distance you want and the smaller sensor camera ends up having more DOF. We don't automatically shoot tight crops with a small sensor cameras just because it makes the lenses "longer".

Thank you for that.

And the 'long lens effect' is also misleading, because the 'look' of (for example) the 85 f/1.2 isn't attained by any 50mm lens just because the sides of the frame (sensor) are lopped off.  This is critical when shooting models in a fashion sense, where a longer focal length is often desirable for compression. Many fashion shooters I know shoot with the 70-200, 200 f/2 or in medium format, a 150.  To get the benefits of those lenses with the crop sensor body, you need to back your butt up even further from the model to achieve similar composition. That can reduce the interaction between model and photographer that can be important to some shoots, requires longer shooting bays (which are often not possible - moving walls and all), and you still don't get the same shallow depth of field that the full frame camera will offer, as noted above.

If you're on a budget, by all means use a crop frame. Amazing pictures can be made with pretty much any camera. But don't argue that it's a better tool for fashion unless you know what you're talking about, and it seems to me, an error like this fundamentally undermines the entire test, and the opinions of the testers.

[bjanes]

Ok ... so now we understand how the random poisson distribution of photons causes shot noise ... but I'm curious about the mechanism to which you are referring.

Photons arrive at the photo site and arouse electrons who flow into capacitor wells when voltage is applied, right?

Poisson distribution causes shot noise because the photons don't all arrive "as expected" ... there is some random fluctuations in their arrival patterns.

Is that the source of PRNU?  

If not ... what is the source of PRNU and what is the impact of fluctuations in the electron activity?

Sorry for my layman's terminology - it has been more than 20 years since I studied physics ...

The photo response and amplification vary among pixels. This noise is usually more of a problem with CMOS than CCD, because in CCD the pixels go through a single amplifier and analog to digital converter unless parallel processing is used. PRNU is discussed briefly by Emil Martinec. One way to estimate PRNU is to take two identical exposures of a uniform target, add a bias to one frame to avoid negative numbers and subtract the frames. One usually takes a 200 x 200 or so sample from the central portion of the image. The resulting standard deviation is the noise for two images, so divide by the square root of two to get the noise for a single image. The subtraction eliminates pattern noise due to PRNU, and PRNU can be determined by subtracting in quadrature (squares of the values) the random noise from the total noise. See Roger Clark for details.

[ErikKaffehr]

Hi,

Thanks for the test, but...

To me this way of testing has a lot of issues.

1) Having a live subject it is very hard to compare between images
2) A lot of discussion about color rendition but how much of that is just a white balance issue? Why not shoot a Whibal or Color Checker Passport.
3) Lot of discussion about sharpening, but D300s image is very soft, is it incorrectly focused or is "uprezzed"?
4) How we do now that proper focus has been achieved?
5) If we discuss sharpening I'd suggest that as much sharpening is done in ACR as possible as capture sharpening in ACR is quite advanced and "unsharp mask" is quite crude.
6) I cannot see the banding in the Nikon samples? Can anyone else?
7) In a sense I don't understand the discussion related to the crop factor. You would choose lenses for the camera not the other way around. Some normal zooms like the 16-85 Nikkor have a better range than a 24-70 on full frame, having one lens instead of two can be nice, of course.

I'm most thankful for all published tests, and my opinion on this test may be a bit harsh, but the idea is that this would be a professional review site so I guess that they would take some criticism.

Best regards
Erik

)