Discrete wavelet transform application in a CNN-based reverse time migration with multiple energy
Shang Huang, Daniel O. Trad
In seismic imaging, image resolution and accuracy are affected by migration approaches. Deep learning has recently been considered an alternative and efficient way to improve image quality. In this project, discrete wavelet transform (DWT) is applied with U-Net on migration data containing multiple energy. The neural network approximates the inverse of the Hessian to obtain high-quality reflectivity prediction. Results show that the DWT subband helps the model learn smooth input, extract critical features from data, and enhance image resolution. Multiple energy provides valuable information for subsurface structure expanding prediction illumination.