Discretization and Filtering Effects on Black Hole Images Obtained with the Event Horizon Telescope

Interferometers, such as the Event Horizon Telescope (EHT), do not directly observe the images of sources but rather measure their Fourier components at discrete spatial frequencies up to a maximum value set by the longest baseline in the array. Construction of images from the Fourier components or analysis of them with high-resolution models requires careful treatment of fine source structure nominally beyond the array resolution. The primary EHT targets, Sgr A* and M87, are expected to have black-hole shadows with sharp edges and strongly filamentary emission from the surrounding plasma on scales much smaller than those probed by the currently largest baselines. We show that for aliasing not to affect images reconstructed with regularized maximum likelihood methods and model images that are directly compared to the data, the sampling of these images (i.e., their pixel spacing) needs to be significantly finer than the scale probed by the largest baseline in the array. Using GRMHD simulations of black-hole images, we estimate the maximum allowable pixel spacing to be approximately equal to (1/8)GM/c^2; for both of the primary EHT targets, this corresponds to an angular pixel size of <0.5 microarcseconds. With aliasing under control, we then advocate use of the second-order Butterworth filter with a cut-off scale equal to the maximum array baseline as optimal for visualizing the reconstructed images. In contrast to the traditional Gaussian filters, this Butterworth filter retains most of the power at the scales probed by the array while suppressing the fine image details for which no data exist.