基于在线字典学习算法的地震数据去噪应用

Application of seismic data denoising based on online dictionary learning algorithm

为了解决常规去噪方法不能根据地震数据自适应构造基函数,去噪效果无法达到最佳的问题,引入基于稀疏表示的在线字典学习(ODL,online dictionary learning)算法对地震数据进行去噪处理。ODL算法能够快速学习,得到与地震数据高度匹配的字典,该自适应字典代替了传统域变换方法中的固定基函数。同时,结合稀疏表示的思想,使用最小角回归(LARS)算法求解出字典的最优稀疏表示系数,将字典与稀疏表示系数组合,从而得到去噪后的地震数据。理论模型和实际地震数据的去噪应用表明:相比较为先进的curvelet变换方法,ODL算法可以更有效地去除随机噪声、相干噪声,同时很好地保留了数据特征。因此,ODL算法对于地震噪声压制有实际指导意义。

Since conventional denoising methods cannot construct basis function adaptively from the data and seismic denoising cannot achieve the best results,an online dictionary learning(ODL)algorithm based on sparse representation is introduced to denoise seismic data.ODL algorithm fast learns a dictionary highly matched seismic data.The adaptive dictionary replaces the fixed basis function in traditional domain transformation methods.Combined with sparse coding strategy,least angle regression(LARS)algorithm is used to solve the optimal sparse representation coefficients of the dictionary.The dictionary and sparse representation coefficients are combined to obtain the denoised seismic data.The application of synthetic models and real seismic data show that ODL algorithm can remove random noise and coherent noise more effectively than state-of-the-art Curvelet transform method.ODL has excellent performance for preserving seismic features and has guiding significance for seismic noise suppression.