Abstract: In the marine environment, the sea clutter has obvious non-stationary and non-Gaussian characteristics. The center frequency of the sea clutter spectrum is not fixed and the spectral width is wide, which seriously affects the radar's detection of the sea surface target. Traditional frequency domain filters are difficult to suppress complex and varied sea clutter effectively. In this paper, by analyzing frequency variation characteristics of sea clutter, we calculate the center frequency smoothness of the spectrum at each distance unit quantitatively and propose an improved time-frequency filtering method based on time-frequency energy distribution for sea clutter. The frequency spectrum dimension of sea clutter is divided into a plurality of frequency intervals, and the frequency range of clutter is determined by determining the frequency interval with the largest energy occupation ratio, and the amplitude mean value of the interval is calculated as the filtering threshold value. Performing filtering performance analysis on actual sea clutter data, the results show that the proposed algorithm has better filtering performance than traditional algorithms and can be applied to clutter suppression under various sea conditions. It is verified that the algorithm can filter out sea clutter signals effectively and protect low-speed target signal energy, improve target signal-to-noise ratio, and play an important role in improving radar detection capability.