Random noise suppression and super-resolution reconstruction algorithm of seismic profile based on GAN

Abstract In this paper, we propose a random noise suppression and super-resolution reconstruction algorithm for seismic profiles based on Generative Adversarial Networks, in anticipation of reducing the influence low resolution profiles. Firstly, used residual learning strategy to construct de-noising subnet accurate separate interference basis protecting effective signal. Furthermore, it iterated back-projection unit complete high-resolution sections image, while responsed sampling error enhance performance algorithm. For data characteristics, designed discriminator be fully convolutional neural network, larger convolution kernels extract features continuously strengthened supervision generator optimization during training process. The results synthetic actual indicated that could improve quality cross-section, make ideal signal-to-noise ratio further reconstructed cross-sectional image. Besides, observations geological structures such as fractures were also clearer.