改进的曲波变换及全变差联合去噪技术

运用常规的基于曲波变换和全变差的联合去噪技术,可以有效地衰减随机噪声,较好地克服使用曲波变换带来的强能量团以及在同相轴边缘产生的不光滑现象,但是这种常规的联合去噪方法对有效信号有一定的损害。笔者采用一种多尺度多方向改进的Donoho阈值去噪思想,较好地克服了常规的联合去噪方法的缺陷,保护了有效信号。该方法在应用曲波变换去噪时,对每一个尺度的每一个方向都选取一个合适的阈值因子,而不是常规的方法对整个曲波系数矩阵只选取一个固定比例的阈值因子。理论模型与实际资料的处理结果表明,该技术最大限度地保留了地震数据的有效信号,在地震资料处理中具有较好的应用前景。

AN IMPROVED RANDOM ATTENUATION METHOD BASED ON CURVELET TRANSFORM AND TOTAL VARIATION

Random noise can be effectively attenuated based on conventional combination of curvelet transform and total variation tech-nology.This combination technology can reduce the pseudo-gibbs effects and the aliased curves resulting from using curvelet transform, but this method is not conducive to the fidelity of seismic data processing.In this paper,a random noise attenuation method is put for-ward based on multi-scale and multi-direction improved Donoho thresholds,This improved combination technology can very effectively overcome the disadvantages of conventional combination technology and better preserve the signal of seismic data. When this method is used to attenuate random noise,we must choose appropriate threshold factors at every scale and in every direction,and it is unlike con-ventional technology which only chooses one fixed proportion threshold factors of all curvelet coefficients.Theoretical model and real data processing results show that this technology can maximally preserve the signal of seismic data,so it has a good prospect in the seismic data processing.