With increasing computing capability, the use of elastic waves in three-dimensional subsurface imaging is becoming more feasible. However, the research in this area is still in the embryonic stage. In conventional seismic imaging, multicomponent data are decomposed into compressional and shear components, and solutions of the acoustic-wave equation are used to process each component independently. Such an imaging technique, however, cannot correctly handle elastic-wave conversions/couplings in complex regions, which is critically important for high-resolution and reliable imaging. In this paper, we develop elastic reverse-time migration imaging, applying a finite-difference solution to the pure elastic-wave equation in heterogeneous media. We implement numerical reverse-time migration imaging in a scheme similar to time-reversal acoustics in the laboratory. To correctly handle polarizations of compressional- and shear-waves during imaging, we also develop a novel vector-imaging condition for elastic-wave reverse-time migration. We use synthetic reflection data to demonstrate that, compared with the conventional imaging condition, our new vector-imaging condition increases image resolution and reduces image artifacts. The new imaging algorithm can significantly improve our capability to image complex subsurface structures.