Abstract: Denoising is an important method in seismic prospecting and has a great impact on the quality of seismic data. Vector decomposition method is to use the angle of noise deviation from the signal to realize suppression of the random noises, belonging to a kind of angle filtering. This method is suitable for prestack and poststack data processing, not restricted by the formation dip, but there is still the problem that noise is not completely separated. This paper proposes a method for further smoothing of the angles by using improved vector decomposition approach based on high?dimensional vector function and spline function to improve the accuracy of the vector angle calculation, in view of the shortage of the discontinuous angles between adjacent tracts after conventional noise suppression. The real seismic data processing shows that this method can be used to get more efficient and more accurate separation between signal and noise, and to filter out random noises, part of multiple and oblique interferences, The case study indicates that this method has good effect of suppressing noise,compared with conventional vector decomposition method