Juan,
I went back and revisited Jordi's presentation, which now makes a lot more sense. However, in doing some experimenting with a stack of 20 calibrated frames, I noticed that the Avg SNR Increment is actually a bit higher for the Windsorized Sigma and Linear Fit, compared to the Average stack with no pixel rejection. I thought that the SNR should be highest with Average/No Rejection, and that the various rejection algorithms could then be tested against their approach to this best value. Am I missing something?
Thanks,
Sean