In the spectral imaging system, random jitter and posture change of the aircraft generated random image motion, and flight of aircraft caused forward image motion. Both of image motion can cause image blur in a longer exposure time, which need for image motion compensation. Due to limited field of view of the optical system, limited size and weight, a stable FSM (Fast Steering Mirror) was used for random image motion compensation and a compensation FSM was used for forward image motion compensation. In the random image motion compensation, inertial sensors were used for measuring the random jitter and the posture change of the aircraft. As the advantages and disadvantages for the gyroscope and inclinometer, we used data fusion of the two sensors to complementary advantages with closed-loop mode filter data based on the frequency domain. In this way, we got high linearity, little drift, high bandwidth and little electrical noise inertial measurement sensors. On the other hand, the motion of the compensation mirror was broken down to the amount of displacement within the time required for each interrupt movement. Under strict timing control, macro forward image motion compensation was realized in the exposure time. The above image motion compensation methods were applied to actual spectral imaging systems, aerial experiment results show that image motion compensation obtained good results and met the remaining image motion compensation image error was not more than 1/3 pixel.