An improved multi-feature fusion scheme is proposed to address the problem of edge signal loss in existing clutter filtering methods for millimeter-wave cloud radar. A discriminative model is constructed based on reflectivity, time and vertical continuity for preliminary clutter identification, and then a morphological binary expansion operation is introduced to generate cloud edge candidate regions, and an accurate edge determination is performed with the help of domain analysis. It is verified that this scheme can effectively filter out clutter while retaining cloud edge signals more completely, solving the problem of edge signal loss in the existing clutter filtering scheme, improving the quality of millimeter-wave cloud radar data, and providing more reliable data support for atmospheric physics research and weather forecasting.