Abstract: Medium frequency radar can detect the wind field in the middle and upper atmosphere, and the full correlation analysis method is one of the common methods for wind field inversion. The objective atmospheric environment interference and the internal interference of the radar system can be equivalent to the additive noise of the receiver, which can lead to the distortion of the shape of the correlation function and affect the accuracy of wind speed estimation, and echo noise reduction is an important signal processing step of medium frequency radar. In this paper, the image processing noise reduction method is introduced into the medium frequency radar, and median filtering based delay correlation parameters (MF-DCP) estimation algorithm. In order to evaluate the performance of the algorithm, the weighted mean square error index of wind speed estimation is proposed, and it is decomposed into the weighted sum of the mean square error of delay correlation parameters. Theoretical analysis and simulation examples show that compared with the existing widely used fitting method, the MF-DCP algorithm can obtain better estimation performance in the low signal-to-noise ratio region, and the influence of the selection of filter window width and hysteresis points on the estimation performance is also given, which provides a reference for the implementation of the MF-DCP algorithm in the engineering practice of medium frequency radar.