After the coefficients are found for Hemispherical Harmonics, how does RTIBuilder determine the direction of the surface normal? Does the program assume the surface is diffuse and look for the maximal lighting direction? I have seen multiple articles (Manfredi et. al.) that state that HSH more precisely determines surface normal direction, but neither those articles nor their references seem to point to how the surface normals are actually calculated. Unlike PTM, there are potentially multiple places where the gradient could be zero and with 16 coefficients, those locations are not easily solved for. Thanks for any assistance!