The utilization of geomagnetic field data especially in the ultra-low frequency range has been shown by prior studies to have the potential to detect earthquake precursors. The most widely used signal processing method for this purpose is arguably the polarization ratio analysis. In this paper, the method was improved by introducing a new normalization process with dissimilar ranges for the vertical and horizontal components. The normalized method was applied to geomagnetic field data that were recorded at locations near the earthquakes which had occurred in Sumatra, Indonesia on 30th September 2009 (M7.6) and Honshu, Japan on 12th May 2015 (M6.8) to evaluate the method’s reliability and effectiveness in two different regions, i.e., the equatorial and mid-latitudinal regions. The results showed that the precursors obtained from the normalized method were less disturbed by random fluctuations and had more distinguishable amplitudes compared to the non-normalized classic method in both studied regions.