It has been demonstrated both theoretically and experimentally that the Green's function between two receivers can be retrieved from the cross-correlation of isotropic noise records. Since surface waves dominate noise records in geophysics, tomographic inversion using noise correlation techniques have been performed from Rayleigh waves so far. However, very few numerical studies implying surface waves have been conducted to confirm the extraction of the true dispersion curves from noise correlation in a complicated soil structure. In this paper, synthetic noise has been generated in a small-scale (<1 km) numerical realistic environment and classical processing techniques are applied to retrieve the phase velocity dispersion curves, first step toward an inversion. We compare results obtained from spatial autocorrelation method (SPAC), high-resolution frequency-wavenumber method (HRFK) and noise correlation slantstack techniques on a 10-sensor array. Two cases are presented in the (1–20 Hz) frequency band that corresponds to an isotropic or a directional noise wavefield. Results show that noise correlation slantstack provides very accurate phase velocity estimates of Rayleigh waves within a wider frequency band than classical techniques and is also suitable for accurately retrieving Love waves dispersion curves.