For SOFI/ISAAC data, the execution times on an UltraSparc or a similar HP station are about 30 seconds per frame (CPU time). A detailed analysis of the procedure with a profiler shows that reducing 64 SOFI frames reaches an order of 1 billion pixel operations to run the whole algorithm. Most time is spent in sorting out pixel lists for non-linear filtering, which has lead to a detailed analysis of fast algorithms for median search and pixel quick sort (see above paragraph about algorithmic optimizations).