A passive stereo vision system for high resolution dense 3D imaging of sea surface is developed. The system is designed for the real-time retrieval of the 3D model of sea surface, in order to extract the spectral features of the wave field to be used for validating and optimizing nearshore wave dynamics numerical models. The system is low cost and easily deployable onboard both fixed and mobile platforms, providing dense coverage of the target areas and fast processing time deriving from a Pattern Recognition (PR) approach. The proposed method for extracting the wave field spectral features brings the computational costs down by remapping each 3D point of the sea surface to an elevation map based on the mean water level. Basic PR techniques are applied to this simplified domain in order to identify crests and troughs of waves and to represent them as oriented lines. Wave typical characteristics, such as significant wave height and mean period, are determined by processing in time and space the identified lines. The system was tested in different acquisition sessions in the Ligurian Sea and it proved to be effective in characterizing the observed wave fields. The use of stereo imagery coupled with PR methods is a promising innovation, as commonly arrays of point measurement devices or the virtual point probe approach in 3D-based techniques are used for the wave field characterization, implying high infrastructural and computational costs.