A two-dimensional truncated singular value decomposition (TSVD) technique is first proposed to enhance spatial resolution of earth observation (EO) products related to ocean-atmosphere interaction. The technique is based on the assumption that the gain function is a two-dimensional tensor product. To set the value of the truncation parameter the Generalized Cross Validation is adopted. Experiments undertaken on a data set of simulated two- dimensional radiometer measurements show the robustness of the technique against the additive noise level and its effectiveness in both reconstruction accuracy and processing time.