Noise suppression test: 
original noise, Poisson distribution, mean = 1599.93, std = 39.9695
original noise, Poisson distribution, mean = 1599.93, std = 39.9695
Scott Lee
shares
Gaussian filter with sigma = 1 pixel 
mean = 1600.02, std = 11.3399
mean = 1600.02, std = 11.3399
Scott Lee
shares
Gaussian filter with sigma = 2 pixel 
mean 1600.01, std = 6.60093
mean 1600.01, std = 6.60093
Scott Lee
shares
Gaussian filter with sigma = 3 pixel 
mean = 1600.00, std = 5.96230
mean = 1600.00, std = 5.96230
Scott Lee
shares
Lee additive noise filter: 
mean = 1599.99, std = 7.47769
mean = 1599.99, std = 7.47769
Scott Lee
shares
Lee multiplied noise filter: 
mean = 1599.97, std = 7.43983
mean = 1599.97, std = 7.43983
Scott Lee
shares
Lee additive & multiplied noise filter: 
mean = 1599.97, std = 7.43983
mean = 1599.97, std = 7.43983
On above test, all filters are using same box width, 7 pixel.
Gaussian filter is faster than Lee's filter, but looks like Lee's filter can preserve more signal information.