Polarization filters have emerged as a useful tool for coherent noise denoising in multicomponent seismic data. By exploiting the polarization properties of seismic waves, these filters can significantly improve wave type separation, which is a crucial step in seismic data processing. However, their effectiveness in land seismic data in complex areas can be impacted by the presence of scattering noise and their applicability limited by irregular spatial sampling, which hinders velocity information for wave type discrimination. To address these limitations, we introduce a novel two-step method for distinguishing and separating polarized wave types in shot gathers captured from a spatially undersampled linear array of 3C vector sensors. The first step involves conducting a polarization analysis in the frequency-slowness domain using elliptical elements derived from the linear Radon transform of the seismic data. The second step isolates specific polarized waves using the 3C Frequency-Slowness Polarization Filter (3C-FSPF), a technique that employs tailored activation functions based on the polarization analysis. Our method stands out from single-station filters that cannot utilize velocity information for wave type discrimination and existing multi-station filters that require regularly sampled data. To evaluate the proposed method, comprehensive tests were conducted on both synthetic and real data. The results consistently demonstrate the effectiveness of 3C-FSPF in separating diverse polarized wave types under conditions of irregular and sparse spatial sampling and noise. Our findings underscore the potential of this method for advancing exploration geophysics by enhancing the quality of seismic data, particularly in land regions with complex near-surface structures.