Contrary to conventional wisdom, higher resolution actually compensates for noise
Sunday, November 16, 2008 News| Introduction | Modeling small pixels | Comparing the SNR of cameras with different resolutions | Real-life comparisons | Conclusion |
Consider the data on 5 full-frame DLSRs shown in the table below. They all have different resolutions. Note that the result for the Nikon D3 puts it far ahead of the other models in terms of RAW noise, followed by the Canon EOS 5D.
The normalized SNR for an equivalent resolution 8Mpix shows the Nikon D3 keeping an easy lead, but the difference between the EOS 1Ds Mark III is no longer huge. Further, the relatively low resolution of the EOS 5D lowers its performance compared to the higher-resolution sensors by the same manufacturer, making the comparison with a normalized resolution more relevant. Indeed, the images are not meant to be visualized at 100% zoom on a regular screen, since only a small part of the image is visible. On the other hand, the images can be printed using different and possibly large formats.
Model |
Release date |
Pixel pitch (µm) |
Resolution (Mpix) |
SNR 18% (effective ISO 200) |
Normalized SNR 18% (effective ISO 200) |
Nikon D3 |
8/23/2007 |
8.5 |
12.2 |
38.99 |
40.82 |
Canon EOS 1Ds Mark III |
8/20/2007 |
6.4 |
21.5 |
35.05 |
39.35 |
Canon EOS 1Ds Mark II |
9/21/2004 |
7.2 |
17.1 |
35.12 |
38.42 |
Canon 1Ds* |
9/4/2002 |
8.8 |
11.1 |
34.72 |
36.14 |
Canon EOS 5D |
8/22/2005 |
8.2 |
13.2 |
35.99 |
38.17 |
*While the Canon 1Ds is far behind, note that it was the very first full-frame DSLR.
Normally, the user cannot access the effective or “real” ISO through the camera settings, but must rely on the ISO announced by the manufacturer, which is usually different. However, DxO Labs has carefully tested the above cameras to compare absolute ISO values with the ISO values announced by the manufacturers (see the results).
Effective ISO values provide the most relevant comparisons: images are shot with the same sensitivity—that is, shadows and highlights are preserved or lost in the same way, thus preserving the ranking when converting the RAW images using the same RAW converter. If we look at the minimal ISO setting (thus the maximal SNR), we see that the lowest ISO setting for Nikon D3 is ISO 200, whereas it is 100 for the Canon DSLRs. The noise measurements at minimal sensitivity are shown below.
Model |
Manufacturer min ISO |
Effective min ISO |
SNR 18% Max |
Normalized SNR 18% Max |
Nikon D3 |
200 |
131 |
41.03 |
42.86 |
Canon EOS 1Ds Mark III |
100 |
70 |
38.58 |
42.89 |
Canon EOS 1Ds Mark II |
100 |
79 |
38.19 |
41.49 |
Canon 1Ds |
100 |
93 |
37.92 |
39.34 |
Canon EOS 5D |
100 |
88 |
38.75 |
40.93 |
In this case, the EOS 1Ds Mark III and D3 are tied. However, the 1Ds Mark III image will be underexposed compared to the D3 (about 1Ev in manual mode), or the exposure time will be twice as large (assuming that auto-exposure strategies are the same).
Further readings for the Contrary to conventional wisdom, higher resolution actually compensates for noise
To provide photographers with a broader perspective about mobiles, lenses and cameras, here are links to articles, reviews, and analyses of photographic equipment produced by DxOMark, renown websites, magazines or blogs.
|
The new "advanced compact" category that adds DSLR features (such as RAW output, manual operation, and interchangeable lenses) to classy, compact-style cameras, continues to be populated with new models—e.g., Canon Powershot G11 and S90, Olympus EP1 and EP2, Panasonic GF1, Leica X1, etc. |
|
On DxOMark, we evaluate and rank many types of digital cameras with image sensors that vary widely in pixel count, pixel size, and digital signal processing. To ensure that sensor performance comparisons between cameras are fair, it is very important both to test under identical shooting conditions and to take viewing conditions into account. |