EOS 5D MkII Noise

You'd think that a noise comparison would be easy. Just take a shot with a 5D and a 5D MkII and see which has most noise. Well, it's not quite that simple. For one thing, when printed at the same ppi (pixels per inch), the image from the 5D MkII would be larger. If printed at the same size, the 5D image would have to be upsized or the 5D MkII image would have to be downsized - and downsizing can lower the noise level.

Then there is the fact that the 5D and 5D MkII apply different degrees of noise reduction to high ISO images. Even with high ISO noise reduction off on the 5D MkII, there is still some default noise reduction applied to JPEGs, and with the 5D (which has no high iso noise reduction function), some noise reduction is applied to JPEGs. You can shoot RAW of course, and with RAW it is possible to turn all noise reduction off, but is that fair? Would you ever actually shoot at high ISO with no noise reduction applied? If you apply some noise reduction how much is resolution compromised? How much noise reduction can you apply to a 5D MkII image before resolution drops lower than a 5D image?

So it turns out to be quite a complex question to answer in a practical sense. With all noise reduction turned off and looking at 5D and 5D MkII images side by side, the intrinsic noise level of the 5D MkII is slightly higher, as you might expect from smaller pixels. However that does not mean that prints made using the 5D MkII won't show lower noise and higher resolution than the same size print made from an image shot with a 5D, assuming an optimum level of noise reduction is applied in each case.

In fact for the most part, prints made from images shot with the EOS 5D MkII will show the same or a lower noise level than the same sized prints made from images shot with the original EOS 5D.

Here are a couple of sets of images showing noise at ISO 3200. These were shot as RAW images and converted to JPEGs using Canon's DPP software with the noise reduction turned off. The first set of images compare a 100% crop from the 5D with a downsized image from the 5D MkII. The second set of images compare a 100% crop from the 5D MkII with an upsized image from the EOS 5D. In each case this means that the two images being compared are equivalent to crops from the same sized print.

On my system (17" diagonal monitor, 1280 pixel horizontal resolution), the above two images represent crops from a 27.3" x 41" print.

On my system (17" diagonal monitor, 1280 pixel horizontal resolution), the above two images represent crops from a print which would be about 35" x 52.5".

I think it can be seen here that the 5D MkII images have somewhat higher resolution and similar or lower noise than the original 5D images. Above ISO 3200 there is no contest of course, since that is the maximum ISO setting of the original EOS 5D. If you underexpose 5D images by 1 stop, shoot RAW and apply +1 EV exposure compensation in DPP you can "simulate" ISO 6400, but the image quality (especially in shadow areas) drops significantly. Below ISO 3200 noise levels are lower and so difference between the 5D and 5D MkII are less visible.

Incidentally, the above images also show that the color rendition of the EOS 5D and 5D MkII are very similar, as you would expect.

The bottom line here is that for all practical purposes, the smaller pixels of the EOS 5D MkII do not lead to higher noise levels in the final prints, and they do give rise to higher resolution images. Again, the differences will be most visible in large prints. If you are only printing small prints you may not see much of a difference between images shot with the 5D and 5D MkII.

The plot below is reproduced (with permission) from dxomark.com who have conducted controlled quantitative scientific tests on many DSLRs. It shows the SNR (signal to noise ratio) in 8x12" prints made from files shot of an uniform 18% grey target with an EOS 5D and an EOS 5D MkII at various ISO settings. These results quantify what I have seen qualitatively, i.e. that the noise level in prints from the EOS 5D MkII are somewhat lower (i.e. the SNR is higher) than that in prints from the original EOS 5D.

Incidentally, this plot also shows that at ISO 3200 (indicated), the true ISO of the 5D is 2700 and the true ISO of the 5D MkII is 2133. This means that the 5D is about 1/4 stop slow and the 5D MkII is about 3/5 stop slow. This in turn means that the 5D is about 1/3 stop more sensitive at an indicated ISO 3200 than the 5D MkII is. This difference holds pretty much constant throughout the common ISO range of the two cameras. Again this corresponds to my observation that at the same shutter speed, aperture and ISO setting, the 5D MkII images are slightly darker than the 5D images.

The 5D MkII offers ISO setting up to 6400, plus H1 (12,800) and H2 (25,600). The higher settings get pretty noisy and so may be of limited use unless you're only making small prints or you're shooting for news. If a UFO lands on the white house lawn and you get a shot at ISO 25600, nobody is going to complain about the noise level!

These images (100% crops of detail of a painted ceramic plate) use the default noise reduction set by the EOS 5D MkII when high ISO noise reduction is disabled. As you can see noise is not visible at ISO 100 and is still very well controlled even at ISO 1600. At ISO 6400 chroma noise (color noise) is apparent and at ISO 25600 the chroma noise and luminance (intensity) noise is very noticeable indeed at this degree of magnification.

The sequence of 100% crops shown below show how fast the noise increases at high ISO settings when all noise reduction is turned off:

Noise Reduction

Finally here is a comparison of various noise reduction settings at ISO 3200 on the 5D and 5D MkII. The target is a section of tree bark (a "real world" subject).Here LNR and CNR are the noise reduction settings for chroma noise reduction and luminance noise reduction respectively used in converting the RAW files to JPEGs using Canon's DPP software. Both LNR and CNR can be set on a 0 to 20 scale. At the top both LNR and CNR are set to 6, which is the corresponds to the "standard" setting for high ISO noise reduction. At "low" the LNR/CNR are 5/6, at "strong" they are 7/7 and when high ISO noise reduction is disabled, the LNR/CNR settings are set to 4/4. The 5D applies LNR/CNR settings of 2/0 at ISO 3200.

As you can see in the top image, the "standard" high ISO noise reduction setting results in a loss of resolution and a "plastic" look. When the LNR/CNR settings are reduced to 2/0, the same as the 5D settings and the 5D image is upsized to match the 5D MkII image, the 5D MkII image looks much sharper, and is sharper than the 5D image. However if a little unsharp masking (USM) is done on the 5D image as is shown in the lower crop, the difference between the 5D MkII and the sharpened 5D image is quite small.

Note that noise reduction is automatically applied to JPEGs, but if you shoot RAW you can chose the level of noise reduction in DPP. By default it will apply whatever settings you had in the camera (off, low, standard or strong), but you can override those setting and chose any value from 0-20.

Here's a table of the DPP LNR/CNR settings which the 5D and 5D MkII apply by default when high iso noise reduction is disabled (or in the case of the 5D, is not available).

Conclusion

The bottom line here is that for all practical purposes, the smaller pixels of the EOS 5D MkII do not lead to higher noise levels in the final prints, and they do give rise to higher resolution images. Again, the differences will be most visible in large prints. If you are only printing small prints you may not see much of a difference between images shot with the 5D and 5D MkII

NEXT: Live View and Video