The processing of seismic data for amplitude inversion often neglects an important issue which could be detrimental to amplitude preservation: the effect of sparse and irregular geometry. The problem is severe in wide-azimuth 3-D surveys acquired with the aim of studying the azimuthal variations in the AVO gradient. We present a new technique for processing wide-azimuth surveys that accurately images 3-D seismic data while preserving relative amplitudes describing the offset and azimuth-dependent reflectivity function. The method focuses on both algorithm accuracy and proper handling of irregular geometry and therefore allows for reliable AVO analysis on migrated data. The technique is based on organization of the data into common-azimuth (CA) and common-offset (CO) cubes and allows interpolation into a regular grid before imaging. The regular CA/CO cubes are then migrated using an efficient true-amplitude migration algorithm. We apply this technique in order to invert for a synthetic reflectivity model simulating a real 3-D wide-azimuth land survey over a fractured reservoir. Results show that, as expected, Kirchhoff migration is very sensitive to uneven sampling, which creates strong amplitude distortions. The azimuth moveout (AMO) transformation succeeded in organizing the data as regular gridded cubes while preserving its amplitude information for imaging. We conclude that interpolating seismic data prior to migration is effective in preserving amplitude information and that AMO can effectively correct for the irregular sampling in wide-azimuth surveys.