Imaging shallow subsurface voids, such as karsts, sinkholes, pinch-outs, dikes, and man-made voids, is an important task in near-surface geophysics. We have developed a new diffraction-based methodology for void detection and imaging. Due to the low signal-to-noise ratio of the diffracted signal in surface acquisition setups, we advocate the use of an SH-wave multicomponent crosshole acquisition. Naturally, the same setup can be used for velocity model building using tomography and for void imaging. The SH-wave data are migrated using a model-based, image-point-dependent automatic muting function that separates direct arrivals from diffracted events in the migration process. For the purpose of location and velocity analysis, data are migrated to the depth imaging offset domain. Only when the velocity model and imaging locations are correct will the diffracted energy be coherently focused to the void location and the diffracted event moveout in the migrated gather will be flat. We found that the received diffracted signal is clearer and has better temporal separation compared with a conventional P-wave crosshole survey. We determined the usefulness of this method using synthetic and field data examples for 2D acquisitions and a synthetic 3D case, showing that a precise imaging is possible. The importance of the S-waves velocity model, which can be extracted from the same survey using conventional tomography methods, is also discussed.