Online Only, you must pre-register to receive access instructions.

NOTE: You Must Be Logged In to Register.

Speaker: Wen Hu, Forland Geophysical Services
Co-Authors: Xianhuai Zhu and Junru Jiao, Forland Geophysical Services

Seismic denoising is one of the key steps in seismic data processing and has been extensively studied by researchers. A well-denoised data set can help interpreters to be able to see “hidden” events, which are often buried by strong noise in raw data. Usually, conventional denoising methods consider reformulating the data into sparse representation by domain transform or dictionary learning. The idea is to get the noise reduced or removed from reconstructed data by performing complicated, but carefully done, operations to separate signal from noise in sparse data representations. Recently neural network-based processing has been proposed as a very efficient alterative to improve traditional seismic processing. However, neural network-based processing only considers one-to-one or pixel-to-pixel type errors and ignores the spatial pattern structure of the data which may lead to overfitting.  Here we propose a new structure-similarity-aware stacked denoising autoencoder network for seismic data noise attenuation.  In our implementation, we pay attention to noise corruption input and also consider missing data. We have successfully applied this new learning-based method to some of the most difficult data areas, where salt related noise is very different from sediment related noise, and where conventional denoising methods have underperformed. Both synthetic and real data examples are tested to demonstrate the effectiveness of the method for various types of noisy input.

Speaker Biography: Wen Hu, Forland Geophysical Services
Wen Hu is a geophysicist with Forland Geophysical Services. She received her MS degree in Geophysics from University of Houston in Dec. 2018 and PhD degree in Materials Science & Engineering from Tongji University in 2009. Her current research interest is focused on machine learning and deep learning for seismic data interpolation and denoising. Wen also has published several SEG meeting abstracts related to implicit seismic modeling and elastic wave separation for TTI medium.

** Access information will be sent to all registrants after registration closes.

Price List: