Hi Geoff,
I am puzzled by the console estimate for SNR (which I assume is signal to noise ratio--what else could it possibly be?). The figure given is an unbelievable 10092. How is this to be interpreted?
This is the SNR function given by
Equation 12 of
ImageIntegration's reference documentation. Note that this is *not* an estimate of the SNR improvement achieved by the integration process with respect to the integration input data. This SNR function is well documented in the literature; see for example
Gonzalez/Woods, page 354. The value of this function for the integrated image is provided for informational purposes. As described in the documentation, this function is not robust and should not be used as a quality estimator, or to compare results with different images.
Another thing that puzzles me is that if I take a straight average of the data in this particular case (no rejection) I get a median noise reduction of 1.855, but if I tweak the rejection parameters I can achieve a median noise reduction of 1.95.
The effective noise reduction function (ENR) is described in the documentation (see the
Quality Assessment section and Equations 40 and 41). Again, this has nothing to do with SNR. ENR is a robust estimate of the relative noise reduction achieved in low-signal regions. As you change rejection parameters, those regions change and hence you obtain different ENR values. Your goal is to maximize ENR while achieving the necessary outlier rejection.
