An improved multi-feature fusion scheme is proposed to address the edge signal loss problem inherent in existing clutter filtering methods for Ka-band millimeter-wave cloud radar. A recognition model is first constructed based on the temporal and vertical continuity of the reflectivity factor of echo signals to perform preliminary clutter identification. Subsequently， morphological binary dilation operations are introduced to generate candidate regions along cloud and fog edges， and neighborhood analysis techniques are employed to achieve precise determination of signal boundaries. The proposed algorithm is validated against co-located lidar observations. Results demonstrate that the scheme effectively suppresses clutter while preserving cloud and fog edge signals with substantially improved completeness， thereby resolving the edge signal loss problem associated with existing clutter filtering approaches and enhancing the overall data quality of millimeter-wave cloud radar.