Obtaining robust phase estimates from phase differences is a problem common to several areas of importance to the optics and signal-processing communities. Specific areas of application include speckle imaging and interferometry, adaptive optics, compensated imaging, and coherent imaging such as synthetic-aperture radar. We derive in a concise form the equations describing the phase-estimation problem, relate these equations to the general form of elliptic partial differential equations, and illustrate results of reconstructions on large M by N grids, using existing, published, and readily available FORTRAN subroutines.