...$n_{\rm max}^g$

In reality, this assumption does not hold, in particular due to undersampling of the PSF and pixel noise in those images from which the kernel model has to be inferred. Since g enters P (Eq. (7)), and P has to be inverted explicitly (Eq. (11)) or implicitly (Eq. (13)), a modest error in the kernel modeling can have severe implications for the quality of the deconvolution results. But this is also true for any other deconvolution method.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...S

The models for both f and g have unit flux, so the surface brightness of the convolved object h depends on $\beta _f$ and $\beta _g$.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.