Deep learning for attenuating random and coherence noise simultaneously

"We introduce deep learning (DL) to three kinds of seismic noise attenuation: random noise, linear noise and multiple. Compared to the traditional seismic noise attenuation algorithms that depend on signal models and corresponding prior assumptions, a deep neural network is trained based on a huge training set, where the inputs are the raw datasets and the corresponding outputs are the desired clean data. Moreover, DL handles three kinds of noise simultaneously instead of sequentially. The DL method achieves satisfying denoising quality with no requirements of (i) accurate modeling of the signal and noise; (ii) optimal parameters tuning. We call it intelligent denoising. We use a convolutional neural network (CNN) as the basic tool for DL and the training set is generated with wave equation for the multiple, and then manually adding random and linear noise. Stochastic gradient descent is used to solve the optimal parameters for the CNN. Numerical results show DL achieves promising performance in synthetic seismic noise attenuation".