To use the FFT algorithm for Fourier inversion, the irregularly sampled visibility needs to be gridded onto a regular grid. This operation requires interpolation to the grid points and then re-sampling the interpolated function. To get better control on the shape of the Dirty Beam and on the signal-to-noise ratio in the map, the visibility is therefore first re-weighted before being gridded.
The visibility function is measured at discrete points which are not assured to be at regular interval along the & axis, therefore griding of the data onto a regular grid is required.
The data is gridded by convolution, where each visibility is convolved with a function and the values at the grid point averaged.