The amplitude values at a certain moment in a seismic record received at a shotpoint and a detection point might come from reflections of various points on an underground ellipsoid, which has the shotpoint and the detection point as its focuses, and the connected line of the shotpoint point and the detection point as its major axis. When stacking is performed, such amplitude values should be added at various points on the ellipsoid instead of being merely added at the CMP. Of course, the amplitude values are different at different points. The utilization of such a stacking method can realize the stacking of common reflection points, and is hence conducive to the stacking image formation and also to the improvement of resolution power and the signal to noise ratio.