Common-image gathers in the dip-angle domain may be computed in relation to wave-equation migration methods, extended by the subsurface offset. They involve the application of a postmigration local Radon transform on the subsurface-offset extended image. In the dip-angle domain, seismic reflections are focused around the specular dip angle of reflection. This focusing distinguishes them from any other event in the image space. We have incorporated the dip-angle information about the presence of specular reflections into the computation of the conventional scattering-angle-dependent reflection coefficient. We have designed a specularity filter in the dip-angle domain based on a local semblance formula that recognizes and passes events associated with specular reflections, while suppressing other sorts of nonspecular signal. The filter is remarkably effective at eliminating either random or coherent noises that contaminates the prestack image. In particular, our dip-angle filter provides a method for the suppression of kinematic artifacts, commonly generated by migration in the subsurface-offset domain. These artifacts are due to an abrupt truncation of the data acquisition geometry on the recording surface. We have studied their appearance and devised an appropriate formation mechanism in the subsurface-offset and scattering-angle domains. The prominent presence of the kinematic artifacts in image gathers usually impairs the quality of the postmigration analysis and decelerates the convergence of wave-equation inversion techniques. We have determined from testing on synthetic and field data that using the proposed dip-angle-domain specularity filter efficiently eliminates the kinematic artifacts in the delivered gathers. We expect involvement of the specularity filter to increase the reliability and quality of the seismic processing chain and provide a faster convergence of iterative methods for seismic inversion.