16 September 2022 in Engineering Applications of Artificial Intelligence
9 February 2021 in IEEE Geoscience and Remote Sensing Letters
In order to remove random noise in seismic data, this letter proposes a seismic data denoising method based on tensor decomposition and total variation (TDTV). Based on the self-similarity of seismic data, this method first groups similar patches into a stack, then utilizes the low-rank tensor approximation strategy to restore the structural effective information of the seismic section. Considering that the approximate seismic section obtained after CANDECOMP/PARAFAC (CP) decomposition and patch aggregation is unsmooth, this letter introduces the total variation (TV) constraint to perform anisotropic diffusion, and protects edge information while smoothing. Finally, the gradient descent method is used to solve the whole model. The TDTV method proposed in this letter can not only effectively denoise synthetic and field seismic section but also restore structural and edge information. Experimental results show that the proposed method outperforms many state-of-the-art denoising methods.
4 August 2021 in IEEE Geoscience and Remote Sensing Letters
Seismic data contain a lot of information in both spatial ( t-x ) and frequency domains. In order to make full use of the effective information in multiple domains, this letter proposes a low-rank tensor minimization method for seismic data denoising. This model first uses variational mode decomposition (VMD) to decompose the seismic data in the frequency domain and constructs a seismic tensor to highlight the frequency information of the seismic data; then to make use of the spatial similarity and frequency correlation of the seismic tensor, a low rank seismic tensor is built through block matching and a low-rank tensor minimization model is established to attenuate the noise. Finally, the denoised seismic tensor is reconstructed into a seismic section. Experimental results show that compared with several denoising methods, the method proposed in this letter can obtain higher signal-to-noise ratio (SNR) and structural similarity (SSIM) and achieve better denoising effects.
27 October 2021 in IEEE Transactions on Geoscience and Remote Sensing
Seismic data denoising is an important part of seismic data processing and has attracted much attention in recent years. With the rapid development of neural networks, convolutional neural network (CNN)-based denoising methods have been widely studied and used in seismic data denoising due to their unique convolutional layer and weight sharing characteristics. However, the existing CNN-based seismic data denoising methods mainly use fixed-size convolution kernels in a certain layer, which forces the different kernels to extract features from areas of the same size and fails to extract features from various granularities. To overcome this limitation, we make full use of the local similarity of seismic sections, use different sizes of convolution kernels to parallelly extract features from different granularities, and propose a multigranularity feature fusion CNN (MFFCNN) method to remove random noise from seismic data. This method uses convolution kernels with different sizes to extract features from various granularities and uses feature fusion structures to fuse the extracted features. Experimental results show that the MFFCNN proposed in this article can better deal with details and texture information than the compared methods.