Preserving the structural and stratigraphic discontinuities or edges is essential in seismic data processing and interpretation. According to several numerical experiments, it is obvious that random noise has a constant spectral density, whereas the structural features vary significantly within different frequency bands, which means that the ratio between the densities of noise and structural features varies significantly in different frequency bands. Therefore, we propose a method called adaptive hybrid diffusion to attenuate random noise, which utilizes a novel adaptive frequency‐based parameter. First, the adaptive hybrid diffusion method decomposes the seismic sections into several band‐limited portions using variational mode decomposition. These portions are called intrinsic mode functions, in which noise and structural energy have distinct differences. Subsequently, utilizing the adaptive frequency‐based parameter, each intrinsic mode function is divided into several monotonous portions that represent the noise or structural area. Afterwards, the total variation and L2 minimization algorithms are utilized separately to suppress the noise in different band‐limited monotonous areas. The algorithms are chosen dynamically, as the portion changes with the change in the adaptive parameter. Finally, these denoised portions are combined to obtain the denoised seismic section. Experimental results on synthetic and field seismic data showed that seismic noise is effectively suppressed by the adaptive hybrid diffusion method, with the edge details of seismic events well preserved.

中文翻译：

---

基于变分模式分解的自适应混合扩散模型用于边缘保持噪声衰减

在地震数据处理和解释中，保留结构和地层间断或边缘至关重要。根据几个数值实验，很明显随机噪声具有恒定的频谱密度，而结构特征在不同频带内变化很大，这意味着噪声密度与结构特征之间的比率在不同频带中变化显着。因此，我们提出了一种称为自适应混合扩散的方法来衰减随机噪声，该方法利用了一种新颖的基于频率的自适应参数。首先，自适应混合扩散方法使用变分模式分解将地震剖面分解为几个带限部分。这些部分称为内在模式函数，其中噪声和结构能具有明显的差异。随后，利用基于频率的自适应参数，将每个固有模式函数分为代表噪声或结构区域的几个单调部分。之后，总变化和L2最小化算法分别用于抑制不同频带受限单调区域中的噪声。当部分随自适应参数的变化而变化时，动态选择算法。最后，将这些去噪的部分组合以获得去噪的地震剖面。综合和野外地震数据的实验结果表明，通过自适应混合扩散方法可以有效地抑制地震噪声，并保留了地震事件的边缘细节。