The dependency of radar backscatter on the surface wind field is investigated using a nautical X-band radar operating at grazing incidence and horizontal polarization in transmit and receive. This resulted in development of an algorithm for wind retrieval from nautical radar image sequences. The algorithm consists of two parts: In the first part, wind directions are extracted from wind-induced streaks that are in line with the mean surface wind direction. These streaks are visible in the temporal integrated radar images at scales between 100 and 500 m. The orientation of the streaks is determined from the local gradients, which are derived from the radar images smoothed and reduced to the appropriate scales. In the second part, wind speeds are derived from the backscatter of the temporal integrated radar image sequence and the radar retrieved wind direction. The dependency of the radar backscatter on the local surface wind and geometry of the radar is parameterized by training a neural network. The algorithm is applied to radar image sequences acquired by a nautical X-band radar mounted aboard an offshore platform in the North Sea. The radar-derived winds are validated by comparison to in situ wind data measured at the platform. The comparison of wind directions resulted in a correlation of 0.99 with a standard deviation of 14.2¿; for wind speeds the correlation is 0.97 with a standard deviation of 0.85 m s-1. In contrast to traditional offshore wind sensors, the retrieval of the wind field from the backscatter of the ocean surface makes the system independent of the sensors motion and installation height and reduces the effects due to platform-induced blockage and turbulence effects.