CMP crosscorrelation (CMPCC) analysis of surface waves enhances lateral resolution of surface wave analyses. We found the technique of window-controlled CMPCC analysis, which applies two kinds of spatial windows to further improve the lateral resolution of CMPCC analysis. First, a spatial weighting function given by the number of crosscorrelation pairs is applied to CMPCC gathers. Because the number of crosscorrelation pairs is concentrated near the CMP, the lateral resolution in extracting dispersion curves on CMPs can be improved. Second, crosscorrelation pairs with longer receiver spacing are excluded to further improve lateral resolution. Although removing crosscorrelation pairs generally decreases the accuracy of phase velocity estimations, the required accuracy to estimate phase velocities is maintained by considering the wavenumber resolution defined for given receiver configurations. When applied to a synthetic data set simulating a laterally heterogeneous structure, window-controlled CMPCC analysis improved the retrieval of the lateral variation in local dispersion curves beneath each CMP. We also applied the method to field seismic data across a major fault. The window-controlled CMPCC analysis improved lateral variations of the inverted S-wave velocity structure without degrading the accuracy of S-wave velocity estimations. We discovered that window-controlled CMPCC analysis is effective in improving lateral resolution of dispersion curve estimations with respect to the original CMPCC analysis.