jats:titleSUMMARY</jats:title> jats:pAmbient seismic noise tomography has, over the last two decades, developed into a well-established tool for imaging seismic properties of the Earth’s crust. Fundamental mode Rayleigh and Love wave phase velocity dispersion curves can be measured from ambient noise cross-correlation functions (CCF) either using a high-frequency approximation theory, or by fitting the spectrum of the CCF to a Bessel function. Here, we advance the latter approach and present an automated algorithm that fits the phase of the Hankel function to the phase of the causal symmetric part of the CCF in order to determine phase velocity curves as continuous functions of frequency. Synthetic tests verify the reliability of the proposed method in the presence of low signal-to-noise ratio (SNR). Moreover, usage of the phase allows for robust phase velocity measurements at longer periods than when using the zero crossings of the Bessel function only and is, therefore, particularly useful at short inter-station distances. In the frequency domain, acceptable bandwidths of smooth phase velocity curves are obtained in an automated procedure using a set of fine-tuned quality criteria. We apply the method to 2.5 yr of continuous waveform data recorded by 58 temporary and permanent broad-band seismic stations in northern Oman. We obtain 1072 and 670 phase velocity curves for Rayleigh and Love waves, respectively, in the period range of 2–40 s. The data are inverted for isotropic and azimuthally anisotropic period-dependent phase velocity maps. Synthetic reconstruction tests show that the phase velocity maps have a lateral resolution of ∼30 km. The results suggest distinctly different middle to lower crustal architecture between the northern and eastern Oman Mountains. Azimuthal anisotropy shows contrasting fast propagation orientations in the shallow and deep crust, which we attribute to stress-induced and structural anisotropy in the upper crust and to lattice-preferred orientation in the lower crust.</jats:p>