Recently, tilted transverse isotropic (TTI) imaging has become a standard practice in deep water Gulf of Mexico (GOM) to resolve the anisotropic effects of wave propagation in salt-withdrawal minibasins. When compared to isotropic and vertical transverse isotropic (VTI) imaging, TTI prestack depth imaging generally provides flatter common image gathers (CIGs) for wide azimuth (WAZ) data, improves image focusing, and significantly reduces well/seismic mis-ties. This anisotropy is thought to arise from the geometry of sedimentation processes, with the "tilt" applied by subsequent tectonic activity. However, the presence of significant tectonic stress or uneven stress can cause fractures in thin-bed layers, which results in additional directional velocity variation for seismic wave propagation, or azimuthal anisotropy around the bed normals. In these cases, the transverse isotropic assumption is insufficient to explain conflicting residual moveouts among CIGs of different azimuths from TTI imaging. A more general anisotropic model, tilted orthorhombic (TOR), is needed to cope with azimuthal velocity variation in these complex geological settings. In this paper, we use a full azimuth (FAZ) data set from Keathley Canyon, Gulf of Mexico, to derive both TTI and TOR models. With the TOR model, we observe improved gather flatness among azimuths and improved structural imaging. We also demonstrate the advantage of FAZ data in detecting azimuthal anisotropy over WAZ data.