Singular Value Decomposition (SVD) is a new and effective method for random noise suppression. The principle of the SVD filtering technique, which enhances the signal to noise ratio of seismic data, is illustrated. Then, by applying time-domain and frequency- domain SVD techniques to suppress random noise generated from a variety of seismic model data, their processing results are compared with each other. The results indicate that the time-domain SVD technique is only able to enhance flat or near flat events. Whenencountering dip events, it is less effective. However, the frequency-domain SVD technique is able to enhance both flat and dip events, and improve the signal to noise ratio of seismic profile. Three simple synthetic seismograms and one real seismic data are used, and the practical effects of these two SVD methods are verified. The results show that the frequency-domain SVD is effective for suppressing the random noise.