We have previously compared both generalized eigensystem (GES) and Tikhonov regularization methods for estimating epicardial potentials from measured body surface potentials. Both of these methods require the choice of a regularization parameter. In this study we compare two methods for choosing this parameter: the Composite Residual Error and Smoothing Operator (CRESO) method, and a new Zero Crossing (ZC) method. We compared both CRESO and ZC methods for zero and first order GES and Tikhonov regularization methods on a swine model and found that the ZC method sometimes produces smaller errors, but only for small noise levels.