Diffusion Spectrum MRI (DSI) has been demonstrated to be capable of defining orientations of intersecting fibers accurately. However, the scanning procedure is overly time-consuming under the current sampling scheme. We developed a reduced encoding scheme of DSI and proved that decreasing the acquisition points to 62% is feasible by diffusion cross-term correction. We calculated the diffusion cross-term from the acquired data in the center of q space by assuming that image corresponding to any diffusion encoding direction suffers b values from diffusion gradients, imaging gradients and the cross-term between diffusion gradients and imaging gradients. The data in outer q space was then corrected with diffusion cross-term and filled by the data acquired with opposite diffusion gradients according to the geometric symmetry of diffusion anisotropy. Both ex-vivo crossing capillaries phantom and ex-vivo rat brain models showed that this method can decrease DSI acquisition time while preserving the patterns and orientations of probability density function.