Natural visibility weighting ($w_i\propto {\sigma }_i^{-2}$)

Independent visibility samples are combined with weights proportional to inverse noise variance. This maximizes the point-source sensitivity for correctly calibrated samples. Dense regions of the sampled Fourier plane receive more total weight; tapering the long baselines can further emphasize extended structure at the cost of angular resolution. A source-template filter instead also weights by predicted source response.