These results are especially useful for comparing ISO noise between these cameras. Read about the Camera ISO noise tests in the help section to learn more about the tests and how they are conducted. A key take-away from that page is that noise reduction is completely off unless otherwise specified.
There are many MB of files required to be downloaded to make all of the links in this review function properly - please be patient while they load.
And again, this test gets us as close as possible to the image quality delivered directly by the sensor. With the 5D III, Canon promised an about-2-stop advantage over the 5D II and has said that much of this improvement is due to sensor design improvements.
Obviously, the 5D III has two higher ISO settings available than the 5D II does, but ... I haven't been able to think of a scenario that the 5D III's extended ISO 51200 and 102400 setting image quality would be acceptable for - even with noise reduction added (below). As usual, the camera manufacturer's highest ISO settings are apparently only for marketing.
On a direct ISO setting comparison basis, the 5D III's noise improvement over the 5D II is not remotely close to two stops at any ISO setting. Honestly, I really I didn’t expect it to be. This technology is not changing that fast – and Canon didn’t release any disruptive technology in this case. Still, better than what was excellent before is positive. I do see some improvement in the 5D III's high ISO noise - especially in the highest comparative ISO settings - the bar has been raised.
Note that, if you can't see a difference in these comparisons, it is unlikely that you will see a difference in your images.
Noise reduction is generally available to any image during post processing but continues to get plenty of in-camera attention from Canon - and in-camera noise reduction is more typically where large high ISO noise improvement figures are derived from.
I wanted to take a closer look at noise reduction in the 5D III. Following are four noise reduction scenarios along with a set of sans noise reduction images (same as shown above).
The "Manual NR in DPP" examples were captured identically to the without noise reduction samples. Noise reduction was then applied in DPP to match the in-camera "Standard" noise reduction settings which are represented next as "NR In-Camera". "Auto DPP NR" shows the amount of noise reduction being automatically applied by DPP when this DPP preference is enabled, which, as seen in the chart below, matches the settings used by the 5D III when "Standard" noise reduction is selected with the JPG image format - "NR In-Cam JPG".
The chart below shows the noise reduction settings use for the NR sample images. NR settings are expressed in Luminance,Chrominance format.
|ISO||JPG In-Camera||RAW In-Camera||RAW DPP Initialized|
The first two example sets below utilize the Standard Picture Style set in DPP (Neutral was used in-camera). The last three examples utilize the Standard Picture Style in-camera.
All examples again use sharpness set to "1" (very low). The last two use in-camera sharpness settings of "1", which convert to Strength:(1), Fineness:(7), Threshold:(7) in DPP (different than DPP's Sharpness = "1").
Hopefully that was not too confusing - or that I can clear all of this up in the summary below the 5D III noise reduction sample images presented here.
Here are some take-aways from this comparison:
Noise reduction can definitely improve image quality at higher noise level ISO settings. But, noise reduction is destructive - especially in terms of image sharpness.
The "Manual NR in DPP" and "NR In-Camera" results appear the same to me - the process of adding noise reduction during post processing does not appear to be disadvantaged.
The last two examples are using the in-camera sharpness setting of "1" (very low) - and both show significant oversharpening - especially at the lower ISO settings. Look for the strong halos around the color blocks and other details. Again, these examples were shot using the low "1" in-camera sharpness setting. The RAW images are sharper than the JPG images and, even with identical noise reduction settings, the JPG images show less noise than the RAW in this comparison.
Basically, you are given the ability to tune your images - and the camera's default settings may not be the best for all situations. My advice is to shoot in RAW and dial in your shots as desired for each situation.
Important is to remember that, when viewing images from any camera, the settings used to produce those images can completely change their appearance. Someone (Canon, Nikon or another individual) had to make choices for the settings selected to capture and/or process those images. Those choices may or may not have been good ones for the specific examples shown.
The 5D Mark III's ISO 51200 and 102400 remain a complete mess even with strong noise reduction applied - You have to be desperate to use these settings. ISO 12800 and 25600 remain very marginal for my uses. I always shoot in the lowest standard ISO setting that will allow me to get my desired shot, but begin to cringe when settings above ISO 3200 must be employed. Your standards and applications may be different.
Here is another comparison example that includes fine details that better-hide high ISO noise.
Notice the 5D III's improvement in high ISO image quality over the 5D II.
Typically, more resolution brings out more details in the fabric - this is especially noticeable at low ISO settings. The 5D III does not have significantly more resolution than Canon's previous resolution leader, the 5D II, but the 5D III results show a noticeable increase in visible fabric detail - indicating that very good pixel-level detail is being delivered by the 5D III.
Back to the Canon EOS 5D Mark III Review.