Not possible to compute SNR single because background is zero in some data

[jcohenadad]

A way to compute noise distribution is to compute noise standard deviation in the background. However, some images are masked out, and the background value is zero, making it impossible to compute noise statistics. Example:

[jcohenadad]

One possibility however would be to create a background mask in the unmasked region, example:

[jcohenadad]

Another workaround, implemented in f50aead, was to get the mask in the white matter and not correct for Rayleigh noise distribution.