摘要
///The authors introduced the multi-scale and multi-direction characteristic of signal and the principles of the second generation of curvelet transform which has the multi-scale and multi-direction characteristic. According to the optimal approximation property of the second curvelet transform about smooth and second-order continuous differentiable singular function, a threshold method of random noise attenuation was proposed based on the second generation of curvelet transform. On basis of theoretical model, the suppressing effect in multi-direction domain (curvelet domain) was discussed, and was compared to that of traditional wavelet transform. Finally, the method was applied on actual data. The result shows that the method sufficiently utilized the difference of directions for signal and noise, which obtained better attenuation result for random noise, efficiently improved the S/N of seismic data and preserved the effective signal in seismic record.