Phase unwrapping and background removal algorithms directly impact quantitative susceptibility maps. Phase processing techniques have been thoroughly studied for brain applications, but accuracy in the presence of large susceptibility and negligible signal, such as bone and air regions, is unknown. The performance of phase processing algorithms was evaluated quantitatively in simulations with a numerical head phantom and qualitatively in vivo in three head datasets. In these experiments, Laplacian-based unwrapping performed poorly. Accurate background removal remains an open question. Results suggest that Quality-Guided unwrapping should be preferred with background removal using Projection onto Dipole Fields.