In the realm of SAR-based sea surface wind field retrieval, the conventional Fast Fourier Transform (FFT)-based spectral analysis method, which relies on a single maximum value, faces challenges in integrating effective regions within the spectrum that contain wind streak features. Consequently, this study proposes an optimized wind direction extraction method based on bivariate polynomial fitting of the spectrum. Through effective denoising and fitting techniques, this method integrates the effective regions with wind streak characteristics in the spectrum and yields more accurate results. Comparative analyses with ECMWF Re-Analysis 5 (ERA5) data and National Data Buoy Center (NDBC) buoy data reveal that the absolute deviations of the retrieved wind direction and wind speed are 3.90° and 2.33 m/s compared with ERA5 data, and 6.82° and 1.19 m/s compared with NDBC data, with reductions of 11.74° and 0.10 m/s in deviations respectively, relative to those between ERA5 and NDBC data. The experimental results indicate that the bivariate polynomial fitting method exhibits excellent wind direction extraction capability and even outperforms ERA5 data in specific regions, thereby contributing to the research on sea surface climate change.