This is the ninth in a series of posts on the Sony a7RIII (and a7RII, for comparison) spatial processing that is invoked when you use a shutter speed of longer than 3.2 seconds. The series starts here.
Bill Claff, who I greatly respect, commented on yesterday’s post to the effect that he would look at a more stringent definition of hot pixels than the 5-sigma that I used.
I reran the analysis of the 16-exposure a7RII 3.2 second series with a 10 standard deviation (10 sigma) threshold for qualifying as a hot pixel.
Of the 10,000 or so hot pixels observed in all 16 exposures, less than 200 occurred in the same place in all 16.
Steeping the criterion up to 20 sigma:
Now we have about the same number of hot pixels that occur in all 16 captures; the higher threshold has not eliminated many of them. But it’s cut way down on the less consistent hot pixels.
If we go all the way to 30 sigma:
Not much difference.
Now let’s look at a 4-second exposure with the same criterion for hotness:
There is no location that was hot for all sixteen images. There is one that was hot for 4 of them and 4 more that were hot in only one capture. The Sony hot-pixel suppression algorithm is suppressing quite well. If it weren’t for the astrophotographic side effect, we might be jumping up and down and cheering, although I still think these things should be done in postproduction.
What if we do pairwise subtraction on the 3.2-second series to emulate LENR?
That eliminates all the hot pixels that appear the same place in all 16 images.
Since the number of fixed hot pixels is so low here — less than 200 out of eleven million — you might say that we have, by narrowing the definition so far, gotten more nearly the expected results, but results that are only marginally relevant to most real photography.