Fun fact: large correlations (hundreds or thousands of samples in 1D, unsure in 2D) can be efficiently computed using FFTs (if the needle is smaller than the ha

Fun fact: large correlations (hundreds or thousands of samples in 1D, unsure in 2D) can be efficiently computed using FFTs (if the needle is smaller than the haystack, you can use chunked algorithms like overlap-add/save).