Multiparameter stacking is an important tool to obtain a first reliable time image of the subsurface. In addition, it provides wavefield attributes, which form the basis for many important applications. The quality of the image and the attribute estimates relies heavily on the accuracy of the traveltime moveout description. The commonly used hyperbolic common-reflection-surface (CRS) operator reduces to the NMO hyperbola in the common-midpoint gather. Its accuracy, however, depends on the curvature of the reflector under consideration. The conventional multifocusing (MF) operator, a time-shifted double-square-root expression, leads to good results for high reflector curvatures and moderate inhomogeneities of the overburden. We used a new implicit CRS formulation that combines the robustness of CRS regarding heterogeneities with the high sensitivity to curvature of the MF approach. It assumes reflectors to be locally circular and can be applied in an iterative fashion. For simple but common acquisition and subsurface configurations, the new traveltime expression reduces to familiar formulas. Quantitative studies revealed that the new operator performs equally well over the full range of curvatures even in the presence of strong heterogeneities, while providing higher accuracy than the conventional CRS and MF methods. In addition, its application resulted in more reliable attribute estimates and an improved time-migrated section. Comparison of stacking and migration results for the complex synthetic Sigsbee 2a data set confirmed the potential of the suggested approach.