The purpose of this project is to generate a more accurate surface model for the skin of the AustinMan and AustinWoman models. The underlying datasets of the AustinMan and AustinWoman models are comprised of a series of cross-sectional anatomical images which are then extruded to create 3-D volumetric pixels (voxels). These voxels, however, result in staircased boundaries which are a poor representation of the actual tissue boundary. When the staircased models are simulated using electromagnetic solvers, the results are not as accurate due to the lower accuracy of the models (i.e., garbage in yields garbage out). Improving the surface models, using surface extraction algorithms, can yield significantly more accurate results. Several surface extraction algorithms were studied, including marching cubes, marching tetrahedral, and surface nets, to determine a suitable algorithm for extracting a high fidelity surface from the voxel dataset. The marching cubes algorithm was chosen to due to its simplicity, robustness, and wide availability. The marching cubes algorithm will initially be implemented in MATLAB to extract surfaces from a sample sphere dataset. After validation, the code will be converted to Fortran so that it can handle the larger AustinMan and AustinWoman datasets. Visualization and verification of the resulting extracted surfaces will be rendered in parallel in ParaView on the Maverick supercomputer at the Texas Advanced Computing Center. The higher accuracy surface meshes will be used by other members of the computational electromagnetics research group to generate more accurate simulations of power absorption from antennas in the human body.