|
Because in-camera noise reduction uses only a single dark frame to do its subtraction, it actually increases the noise level in the dark areas of the resulting image. To have a significant improvement, you would need to median combine a large number (8-16+) of dark frames (shot at the same temperature as your lights), and then subtract those from your light frames, which need to be shot with noise reduction off.
There are several kinds of noise that you are attempting to remove by using dark frames. Hot pixels are non-linearly responding pixels which, due to manufacturing defects, saturate very quickly during an exposure. There will be a relatively low number of these pixels in the image. Beyond a certain exposure time, these pixels will be 100% saturated, and are removed essentially by doing an average of the surrounding pixels.
The second type of noise which the dark frames address is amplifier glow. This is thermal noise caused by heating of the sensor array due to its proximity to warm circuitry. Some cameras do not exhibit this to a large extent, while others are plagued by it. Essentially it creates an area of the image that is offset due to false signal caused by the heat.
The third type of noise which dark frames can directly help reduce is bias current. This is the base, or offset signal, which the camera uses to help pull the signal up above the noise threshold. Light frames and dark frames will both contain this offset, so when the dark frame is subtracted, the bias is removed as well.
The fourth type of noise which you can partially remove using dark frames is plain old thermal noise. This type of noise is a product of sensor temperature and exposure time, but is essentially a random function and as such, can't be cleanly removed using a single dark frame. This is where in camera noise reduction won't help you. You will end up with a lot of "salt and pepper" noise in your image due to this noise. The more dark frames you are combining to use for subtraction, the more this noise averages out, and the less statistical error there will be in the data. There are formulas for calculating the amount of noise in the dark frame, but basically it comes down to the noise amount reducing with the square root of the number of exposures. IE, 16 frames will have half the noise error that 4 frames will have.
To understand where dark frames can actually add noise to the resulting image, consider that you are stacking your light frames, and this results in averaged data. This averaging has removed some of the error caused by photons arriving at your sensor in a random pattern, resulting in lower noise images. However, if you then go and subtract a single dark frame from it, that dark frame has not had the benefit of noise reduction due to stacking. The resulting data will contain more error after subtraction than it did before, despite the hot pixels being gone. This is likely what you are reporting in your results.
The above are why you hear of people compiling "master darks" and "dark libraries", because it can take a substantial amount of time to shoot a significant number of dark frames at a given temperature, especially as exposure times increase.
Hope this helps,
Eric
|