In recent years, significant progress has been made in the area of electroencephalography (EEG) source imaging. Source localization on simple spherical models has become increasingly efficient, with consistently reported accuracy of within 5 mm. In contrast, source localization on realistic head models remains slow, with subcentimeter accuracy being the exception rather than the norm. A primary reason for this discrepancy is that most source imaging techniques are based on lead fields. While the lead field for simplified geometries can be easily computed analytically, an efficient method for computing realistic domain lead fields has, until now, remained elusive. In this paper, we propose two efficient methods for computing realistic EEG lead-field bases: the first is element oriented and the second is node oriented. We compare these two bases, discuss how they can be used to apply recent source imaging methods to realistic models, and report timings for constructing the bases.