Harmonic phase (HARP) is an image processing technique of MR-tagged images in frequency domain developed to evaluate the regional tissue strain. The filtering stage is a key point in this method because all information about the deformation have to be included in the analysis. In this paper we develop an automatic method to define the filter parameters in order to minimize errors in the strain evaluation and to avoid operator related variability. A software model is built to investigate the error in strain estimation caused by filter parameters variability and to evaluate the developed method. The automatic technique models each spectral peak with a bidimensional Gaussian function and uses these information to create an optimal elliptic filter. We validated the automatic method on simulated data set. In real tagged MR images analysis, the method achieved the same performance as the manual analysis, considered the gold standard, eliminating the inter-operator variability.