My understanding of the in-camera noise-reduction is that it removes the dark current and bias signal noise from images (eg repeatable hot pixels and the signal readout noise of the sensor/electronics).
To do the dark current noise reduction, the camera needs to take an image of the same duration as the actual exposure you want, but with the shutter closed, then it subtracts that signal from your exposure. So if you take a 30 second exposure, it'll take your 30 sec shot, then take 30 secs more for the dark, then it'll save to your card. Because the dark current also contains the bias signal, it gets removed at the same time by default.
You can't do that in post-processing. It needs to be captured in-camera. And I could be wrong, but I'm pretty sure it's applied to the RAW and JPGs.
There are other types and sources of noise, the above only removes those types mentioned. If your exposures are short enough, the dark noise probably isn't that much of an issue. We are more aware of it with astro-imaging because we have extremely long exposures of very dark subjects so signal to noise ratios is much, much, much lower than for "normal" photography.
For all other types of noise reduction, for normal photography, your typical NR software is the only way to go. They'll probably handle the dark and bias as well, but I reckon the best results are to let the camera do that work for you.