K10D Noise Reduction

Can anyone explain how the NR setting on the K10D does its job? Does it use a second exposure (blank) to subtract from the original?

Can anyone explain what happens. I tried it briefly once when I first got the K10D and it appeared to work, ie the image was much smoother but the test was inconclusive.

Soes any one KNOW.
Ron

It's called dark frame subtraction. When NR is active and the shutter speed is 1/4s or slower a second exposure is made for the same duration as the desired exposure this time with the shutter closed. This effectively produces an exposure which contains information relating to only the self excitation of the sensor (stuck or hot pixels) for the given exposure duration. This second exposure is then subtracted from the actual exposure, the net effect of which is to remove static sensor noise from the image. However the random background image noise will generally be a little worse.

Thank you for your response. Have you reached any conclusion as to the effectiveness of this setting? Some early posts to dpreviews seemed to be contrary to my understanding of the process.
Regards
Ron McDermott

Dunno why, but I have yet to find the need to turn on the NR setting on the K10D. (I used to use it all the time on the istD).

Could be that I shoot RAW lately, but I've taken shots with 15 second exposures, and longer, and have not needed noise reduction. (some touch-ups of a few hot pixels) but I was truly amazed at the clear, smooth image that resulted. It could also be the nice job that SilkyPix does on RAW images (default settings).

I'm not sure the NR is worth the extra time between shots to make the dark frame subtraction. I guess I need to do some side by side tests... Too many factors changed in my process since I got my K10D.

Silkypix does an excellent job when it comes to removing hot pixels.
I have the noise reduction in the camera allways turned off.

Hi Ron,

The effectiveness of the in camera NR really depends upon several factors. Firstly if you are shooting jpg exclusively post processing NR is only so capable and can't replicate the effectiveness of real time NR, much like attempting to digitally replicate the effects of optical IR or polarizing filter in post precessing.

When shooting RAW there is definitely more flexibility as many RAW converters have excellent hot pixel removal algorithms, that said they are not 100% effective as they only use clever logic to determine which pixels are hot or stuck. For instance some clusters can be a problem for RAW converters where they are generally completely removed using dark frame subtraction (either in camera or shooting a manual dark frame and performing the subtracting in post processing). The penalty is of course that you always double your effective exposure time which for long exposures can be a pain.

Cheers,

Hi all
Your thoughts have been extremely helpful. I actually shoot RAW (PEF) exclusively and have been using ACR 4.2 to process images which I save as DNG. I find a need to produce image files (JPG) to male Nero slide shows or PSD or TIFF for printing. After a HQ JPG will be 6+MB and the DNG is 9-10.5 MB.
I have been experimenting with HDR and love some of the images. As you will know if you have used Photomatrix the grain can become a problem in dark areas.
Hence my interest in NR.
Regards
Ron McDermott

This is a different issue altogether

The NR setting, as mentioned, is only for long duration exposures. The dark frame subtraction does help - even in RAW. However, the K10D is very, very good (especially compared to the *ist-D). For example, the quality of batteries made a big difference to the amount of CCD noise at long exposures. With the LiIon battery of the K10D, and presumably better electronics too, the problems are a lot smaller. Silkypix does a very good job of isolating hot pixels (Capture One does a passable job, and ACR doesn't do a good job AFAIK).

Regarding the dark areas... the problem is that the CCD is linear, not logarithmic. Consequently, in a 12-bit RAW file there are 2048 values for the highest 1-stop, 1024 for the next, 512 for the next, 256 for the next etc. So, by the time you get to 5-stops under the brightest, you've only got a measly 128 values

Best approach for HDR work:
* Use the auto bracket on the K10D and set 2-stop interval for 5 exposures. This will give you a good range.
* Definitely use a tripod
* Process the RAW files in Silkypix first into 16-bit TIFF files. Silkypix has excellent noise reduction properties. If not, use Capture One LE (free with Sandisk Extreme III cards). Just don't use a curve or contrast adjustment. Process them flat after apply white balance and film type.
* Bring the TIFF files into your HDR app (I use Artizen because it has far more choices when it comes to tone mapping operators. Photomatix uses the Fattal operator which can be a bit "cartoony", if you know what I mean)

As soon as you start trying to pull the shadows too far from the gloom, you'll get noise. However, as I mentioned, a carefully setup Silkypix will give you great HDR results (You'll get the problems you mentioned if you try and push a single RAW file too hard)

Hope that helps!
Matt

I would add that the only way that you can auto-bracket (using a remote so as not to move the camera) with a fixed aperture (obviously varying DOF frame to frame is undesirable in the production of HDR shots) is to set the mode to Av. You can add exposure compensation to trim exposure or pre-set the exposure using the AE-lock.

Good point Rob
The times I did HDR, I used Av mode naturally. As you say, it's vitally important that the aperture does not change throughout the exposure set.

Matt