After clicking “Watch Now” you will be prompted to login or join.
Seismic Image Compression with Deep Learning
Pedro Mario Cruz e Silva, NVIDIA | Jo�o Paulo Navarro de Oliveira, NVIDIA
GTC 2020
The resolution of acquisition sensors is growing exponentially, so we need to rethink the future of very large seismic surveys storage. We'll discuss a deep-learning approach for very low bit rate seismic data compression. Our goal is to preserve perceptual and numerical aspects of the seismic signal while achieving high compression rates. The method is suitable for both pre-stacked and post-stacked seismic images, using 2D and 3D convolutions. It benefits from the intrinsic characteristic of deep learning methods and automatically captures the most relevant features of seismic data. We'll present an approach for training different seismic surveys and perform validation experiments in real seismic datasets, showing that our data-driven approach successfully yields compression rates up to 70:1 with an average peak signal-to-noise ratio around 50 dB.