This study deals with surface waves extracted from microseismic noise in the (0.1–0.2 Hz) frequency band with passive seismic-correlation techniques. For directive noise, we explore the concept of passive seismic-noise tomography performed on three-component sensors from a dense seismic network. From the nine-component correlation tensor, a rotation algorithm is introduced that forces each station pair to re-align in the noise direction, a necessary condition to extract unbiased traveltime from passive seismic processing. After rotation is performed, the new correlation tensor exhibits a surface wave tensor from which Rayleigh and Love waves can be separately extracted for tomography inversion. Methodological aspects are presented and illustrated with group-speed maps for Rayleigh and Love waves and ellipticity measurements made on the San Andreas Fault in the Parkfield area, California, USA.