Abstract: Weak seismic diffractions generated by small-scale discontinuities—such as faults, stratigraphic pinch-outs, and fractured-vuggy systems—contain critical information for high-resolution reservoir characterization and fault identification. However, extracting these weak signals remains challenging because they are routinely overwhelmed by strong specular reflections, migration smiles, and random noise in conventional migrated sections. To enhance the imaging quality of weak diffractions in complex tectonic regions, we propose a dip–coherence constrained Gaussian beam diffraction migration approach. First, ray propagation vectors at image points are rapidly and accurately evaluated from Gaussian beam propagation parameters. These vectors are then combined with local structural normals to construct anti-stationary phase filter weights, which achieve the initial attenuation of reflection energy. Using the omnidirectional distribution of diffractions in the dip domain, we further decompose the filtered diffraction image into two directional components based on the azimuthal relationship between the imaging vector and the structural normal. A novel imaging condition weighted by dip-coherence is subsequently constructed. This strategy effectively identifies and suppresses residual reflections and coherent noise present only in specific directions, thereby enhancing dual-sided symmetric responses while suppressing unilateral residual reflections and migration smiles. Comparative diffraction imaging tests on synthetic models and field seismic data effectively confirm the correctness and validity of the proposed method.