In the theoretical calculation of atmospheric wind fields using the cross-correlation analysis method of Medium-Frequency radar, it is necessary to compute a series of correlation parameters from the received echo signals, such as autocorrelation and cross-correlation functions, within the main lobe range of the antenna array to retrieve atmospheric parameters. However, both theoretical analysis and practical applications have shown that the shape of correlation functions can be affected by atmospheric conditions and receiver noise, leading to significant biases in the estimated correlation parameters within the main lobe range. In this study, we theoretically analyze the influence of noise on the amplitude of autocorrelation and cross-correlation functions. We propose a noise reduction method based on the characteristics of correlation functions at the zero-delay point to calculate the noise factor and process the correlation functions within the main lobe range. Furthermore, we conduct simulation analysis to evaluate the performance of this noise reduction method and summarize the effects of the number of fitting points and fitting methods on the noise reduction performance.