Preprocessing: Stacking

There are many stacking methods and, as you might expect, almost no agreement on which one is best. Allow me to add to the noise on this issue.

To help tighten things up a bit I'm going to refer to cosmic ray, meteor, satellite, airplane, and even firefly tracks in images as transients.

Averaging

This is the method of choice when you have fewer than 8 or so light frames. It won't get rid of transients, but it will give you the best signal to noise ratio. If you don't have any transients, averaging can be a good choice regardless of the frame count.

Median

This should be used only when you need to eliminate transients and have a large number of light frames. Other outlier culling methods do this, too, and seem to impose less penalty on your signal strength.

I use median stacking for non-light calibration frames where I typically have 30 frames of each to combine. Using a sigma clipping stack would also make sense. I prefer some sort of outlier rejection for calibration frames to help reject possible cosmic ray tracks.

Min Max Excluded averaging

Much like median, but can do better in the reduction of noise. You lose the rough equivalent of two light frames using this.

Sigma Clipping averaging

The option of choice for many imagers with the rejection set at anywhere between 2 and 2.5 sigma (standard deviations). At 2.5 sigma you reject about 1.2% of your data; at 2.0 you reject about 4.6% (one frame in twenty)

Table of contents Previous Page Next Page Last Page