I was getting some really wild estimates of sensor gain and read noise for my Canon 6D. After looking carefully at the script, and also doing things by hand in PI with the Statistics tool, it became clear to me that the use of Mean and Standard Deviations were likely culprits in the CCDParameters tool.
So I made the changes to the script to make use of Math.median() and Math.MAD(), and take care to reject pixels at 0.0 and 1.0 in the analysis.
I also have bizarre dark statistics from the Canon, as they appear to be subtracting an internal dark estimate from each image frame, as well as possibly rescaling. I can't do anything about the scaling (yet), but I can compute a dark current from the MAD in a pair of subtracted dark frames. So I added that bit to the derived quantities in the report output.
Here is the modified script. Very slight modifications with original code still intact, but commented around. (Had to change the name to CCDParameters.js.txt to get it to upload here.)
