Seismic Data Compression Using Deep Learning

The exponential growth of the size seismic data recorded in surveys and real time monitoring makes compression strongly demanded. Furthermore, will lead to an efficient use bandwidth assigned for communication link between stations main center. In this paper, two convolutional autoencoders (CAEs) are proposed compression. algorithms mainly based on neural network (CNN), which has capability compress into feature representations, thereby allowing decoder perfectly reconstruct input data. results show that first model is at low ratios (CRs), while second improves signal-to-noise ratio (SNR) from about 3 dB 12 compared other benchmark moderate high CRs.