Yesterday I did not shoot anything but some test images, none of them did I want to make Image of the Day. This is from the archives, an image from the crypt below the Dome of Speyer, Germany. I shot it with the D200 in July 2007, using the Nikon 18-200 VR at f8, ISO 1600 and 1/3s, handheld. I’ve converted it with DxO and added some local contrast later in Photoshop.
I am still in Carinthia, it’s Sunday now and I have managed to install DxO 5.3 yesterday evening. We can carry on with the review.
Let me begin with a correction. In “741 – Just Another Day On Earth” I originally said that “what Adobe Camera RAW does and DxO seemingly not, is the automatic elimination of hot pixels“. This is wrong. Yesterday I found out that DxO can automatically eliminate what they term “dead pixels”, it’s only not on by default, and it is hidden down in the options for noise reduction. Sorry for the false alarm.
In a comment to that post Nick Jungels said that
“Looking at DxO Optics, the noise performance seems very, very similar to running Noise Ninja. At least in my one comparison the pictures were virtually identical (as far as noise goes).
Have you compared the DxO versus any of these other noise reduction options?
At that moment I had not, this is what I want to look into today, and here is the image that we will look at in some variations and detail. You see now why I used an old image for the Image of the Day?
I made three versions of this image: At ISO6400 and correctly lit, at ISO 3200 and one stop underexposed, and finally at ISO 3200 and underexposed by three stops. All three images were shot handheld with the D300 and the Sigma 50/1.4 at f4. High ISO noise reduction in the camera was set to “low” (from a standard of “normal”), exposure times were 1/60s and 1/250s. Fairly normal values for low light street photography. Let’s begin with a 100%crop:
You really have to click on the image for the 100% view to see the differences. Left is the result of Adobe Camera RAW 4.6, in the middle the same with Noise Ninja applied, and on the right is the output of DxO.
As to processing: First I have created the DxO image on standard settings, with the preset for high-ISO images applied. This produces stronger contrast, dark shadows, uses stronger noise reduction and also takes care of dead pixels. Then I have loaded the original RAW file into Adobe Camera RAW, without sharpening and with the exposure parameters set to automatic. Later in Photoshop I used adjustment layers to match this much lighter image to the look of the output of DxO. Basically I pulled all there was into Photoshop, and then toned it down, along with all noise. I think this is fair and as comparable as it gets.
There are quite some differences. Obviously the quality of Adobe Camera RAW without noise reduction is not really acceptable. There is a high level of noise, grain is coarse and we see much color noise as well.
Applying Noise Ninja made the image much better. This is not the detail that we expect from a 12 megapixel image, but for ISO 6400 it is the best that we can get without inventing detail.
DxO delivers a tad more sharpness, but that was to be expected as it did some sharpening and Adobe Camera RAW was set to do not. What’s obviously better, is color noise. Noise Ninja took care of the high-frequency color noise, but some low frequency noise, i.e. big color blotches, remained. The difference is not dramatic, but it is there.
Looking at the differences at 200% is more revealing. DxO produces a very fine grain on pixel level, Noise Ninja struggles with color noise. Overall I’d say that the result of DxO is more pleasing, looks less like digital artifacts, but you really have to care about pixel peeping to get a kick out of it.
Let’s do the same once again, this time with an image taken at ISO 3200, but underexposed by one stop. At 100% we see basically the same situation.
Please ignore the systematic tonal differences and the color difference in the dark tones. That’s an artifact of my method. I simply was not able to exactly match outputs. Let’s concentrate on noise instead.
Noise levels are slightly lower, the difference between Adobe Camera RAW + Noise Ninja and DxO is hardly visible.
At 200% it’s pretty much a tie, but from the looks I’d prefer DxO again. The fine grain actually looks nice, not like typical digital noise at all. On the other hand: both are clearly acceptable.
Then I turned to the ISO 3200 image underexposed by three stops, effectively pushing the image to ISO 25600, and this is the point where DxO breaks spectacularly. I did not even include a 200% comparison here, it is obvious. DxO smudges away detail like mad. I have tried for quite some time to find settings that would improve the result, but to no avail. In fact, this seems to be the best it can do under these conditions.
And Noise Ninja? Still riding the waves. The result that it produces is unusable as well, but compared to DxO it does not break, it degenerates slowly.
What’s the verdict now? If forensics is your job and you need to get the last information out of an image, even if you won’t be able to use it for any aesthetic purposes, then Noise Ninja is your tool. In any other case DxO seems as good or better, partly depending upon your aesthetic preferences.
In my eyes this result is not disappointing, not at all. After all, Noise Ninja is one of the best noise reduction plugins on the planet, one of three or four tools that constitute the state of the art. That DxO plays in that league and maybe slightly tops it, is quite remarkable, given that noise reduction is only one card in DxO’s game.
What about the price? Currently Noise Ninja and Neat Image cost around $70, Nik Software’s Define 2.0 is sold by Amazon for around $80, while DxO Optics will cost you $170 or $300, depending on your camera. Thus you’ll clearly have to need something else out of DxO’s portfolio to justify the purchase.
We’ll look closer into these things in other posts. Stay tuned.
Sorry, no sound samples, but looking for Wendy James on YouTube will give you some videos and an idea of what to expect.