08-16-2016 10:48 AM - edited 08-16-2016 10:49 AM
OK, I thought if we would sweep an array subset across all locations and take a histogram of all subset maxima locations, only the real maxima will have a significant count. Seems to work just fine, try it!
(would need small adjustments for non-square inputs, but the idea stays the same)
08-17-2016 06:58 AM
Thanks for your solution!