Abstract: Anti-stationary phase filtering (ASPF) is a dominant wavefield separation technique for diffraction imaging. By leveraging the wide-angle scattering characteristics of diffractions, ASPF isolates diffraction energy through the suppression of specular reflections within the dip domain. However, the inherently weak diffraction signals are frequently masked by residual reflection energy and migration artifacts (such as migration swings) that conventional ASPF fails to eliminate completely, thereby severely compromising the interpretability of the imaging results. To address these challenges, we propose a dip-domain cross-constrained Gaussian beam diffraction migration method. Utilizing Gaussian beams as robust wavefield propagators enables the efficient and accurate calculation of subsurface ray propagation vectors, establishing a solid foundation for the separation of reflections and diffractions. Furthermore, by exploiting the distinct wide-angular distribution of diffractions in the dip domain, we construct a cross-constraint strategy applied to the decomposed imaging components. This strategy effectively identifies and attenuates residual reflections and coherent noise concentrated in specific directions, significantly enhancing the signal-to-noise ratio (SNR) and focusing of the diffraction image. Comparative tests on both synthetic models and field seismic data demonstrate the validity and efficacy of the proposed method.