Patient-specific computer models of the human atria have the potential to aid clinical intervention in the treatment of cardiac arrhythmias. However, quantifying and integrating the heterogeneous qualities of the myocardium through imaging is particularly challenging due to the unknown relationship between voxel intensity and tissue conductivities. We establish a method to determine the relationship between local conduction velocity and scar density, extracted through the analysis of late-gadolinium enhanced magnetic resonance imaging data.