> You would be far less able to recover individual strands of hair in an image that was gaussian blurred.

Frequency-domain deconvolution is frequency-domain deconvolution, right? It doesn’t really matter what your kernel is.