Abstract: This paper presents an algorithm for retrieving sea surface wind direction, addressing the limitations of existing methods in the spatial domain of marine radar images. Initially, it categorizes all pixel points based on wind streak features in polar coordinate sea surface static feature images. Subsequently, the Gaussian filtering method is applied to identify pixels with the main trend and redistribute them. Finally, employing the gray co-occurrence matrix for angle estimation, the algorithm determines sea surface wind direction based on the relationship between small-scale wind streak and sea surface wind direction. The spatial distribution change in the gray co-occurrence matrix automatically resolves the 180° ambiguity problem in sea surface wind direction inversion. Experimental validation using measured ocean data from the East China Sea demonstrates the effectiveness of proposed algorithm, with a correlation coefficient of 0.93 and a root mean square error of only 6.05°. Furthermore, the algorithm proves capable of mitigating the impact of rainfall on the wind direction retrieving process, confirming its strong engineering application potential.