I've mentioned before that I went ahead and bought a 1Ds3 in December because I thought it would probably compete with the Nikon D3 at ISOs above 3200 once noise reduction had been applied. Having used it for 6 months for shooting extremely low light, I am fairly happy with that decision. While it's useful to shoot JPEG, when you are stretching the limits of high ISO I imagine it's better to shoot raw, even on the D3, so being forced to shoot raw to get these speeds on the 1Ds3 isn't a deal breaker, unless you are a very hurried press photographer.
Here's an example of ISO 10,000 (10K) and an unscaled crop WITHOUT NOISE REDUCTION OR SHARPENING. To get this speed I shot at H (3200) and then pushed by a stop and two thirds in DPP. The shot was taken in a gloomy music studio which was mostly too dark for the camera to auto-focus.
Now of course, the D3 is cheaper, but then again it doesn't quite give you the same detail or cropability at lower ISOs that the 1Ds3 can, provided it is used carefully with adequate glass.
Overall then, from what I've seen, the D3 isn't significantly better, if at all, at very high ISOs, except that your NR is done in camera and you can produce JPEGs more quickly at those speeds.
Richard,
That's impressive alright but since we have several excellent noise reduction softwares, what if you apply some kind of noise reduction work there? I am curious whether the details can still be retained while the noise is reduced. What I am trying to say is if you could apply the proper or necessary post processing adjustment, including some noise reduction adjustments, for us to take a look at?
Good performance. A little chrominance NR would kill the purple noise and you'd likely notice no other difference.
What I'm seeing is that the noise patterns look a lot less 'digital' than what I'm used to on previous Canon models - i.e., push a 5D 3200 hard and you get uneven noise and quite often banding. From your experience, is this pretty much true universally?
I don't see any need for noise reduction here. The picture will get used on a screen at low rez, and on album artwork maybe, and just possibly I might do a nice big print. In all of those applications you won't see problematic noise. The grain effect is beautiful though, as you can see in this web-sized version.
I do get banding in shadows, especially in the blue channel, so it very much depends on the lighting. Obviously tungsten and candles provide very low blue light levels which is probably why most often I see banding in blue. Banding is pretty much impossible to fix. The important thing is that I am seeing between 1 and 2 stops better than the 5D in very low light, so I have no D3 envy.
Well, you might give it a try anyways. I opened your full-sized sample as RAW in Photoshop and full Chrominance reduction. IMHO, it makes the noise look even closer to film grain and de-emphasizes the mild banding.
I'm imagining you're using DPP for the conversion. I'm finding the high-ISO results of ACR more pleasing if you're not in a hurry - it's not as sharp, but the blue channel noise is a lot easier to minimize. I've often used 5D ISO 3200 shots for magazine covers and other reasonably-sized prints.
Here is a shot @ ISO 6400 shot with Canon 5D that I received yesterday. Shot was handheld @ 1/50th of a second, 85 f1.8 @ f1.8 so some motion blur/OOF areas may be there. compared to 20D I am very impressed, not just by the noise level, but the character of the noise. I had looked at the comparable shots between 5D & D3 on dpr, and was not expecting this type of performance at 6400.
What was the actual amount of available light? approx. value in EV is ok.
I have seen a lot of people posted high ISO images, but without knowing the amount
of available light, it is kind of pointless because I can shoot something under sunny sky
or turn on all my bathroom lights (which is way brighter than necessary to groom
oneself) and get cleaner images than yours at ISO 10K
Aperture was either f1.2 or f2 (I can't tell since not a Canon lens), shutter speed 1/500th, iso 3200 pushed 5/3 stops. I needed the speed since he was moving quite a lot. Levels were above candle-light, but as I said, very gloomy. Obviously in candle-light to get this quality one would be restricted to slower speeds.
Sorry, didn't see the word gloomy in your post. My bad
Just curious what the actual scene look like (the actual
available light). Could you also post an image of what the
actual scene should look like since the one you posted has
been pushed one and two-third stops higher (assuming from
actual exposure).
