A practical method is suggested to quantitatively retrieve directional spectra of ocean surface waves from high-resolution satellite sun glitter imagery (SSGI). The method builds on direct determination of the imaging transfer function from the large-scale smoothed shape of sun glitter. Observed brightness modulations are then converted into sea surface elevations to perform directional spectral analysis. The method is applied to the Copernicus Sentinel-2 Multi-Spectral Instrument (MSI) measurements. Owing to the specific instrumental configuration of MSI (which has a primary mission dedicated to mapping of land surfaces), a physical angular difference between channel detectors on the instrument focal plane array can be used to efficiently determine the surface brightness gradients in two directions, i.e., in sensor zenith and azimuthal directions. In addition, the detector configuration of MSI means that a small temporal lag between channel acquisitions exists. This feature can be exploited to detect surface waves and infer their space-time characteristics using cross-channel correlation. We demonstrate how this can be used to remove directional ambiguity in 2-D detected wave spectra and to obtain information describing local dispersion relation of surface waves. Directional spectra derived from Sentinel-2 MSI SSGI are compared with in situ buoy measurements. We report an encouraging agreement between SSGI-derived wave spectra and in situ measurements.