Enhancement of Seismic Image Based on Novel Approach with Custom U-Net Architecture

This paper introduces an innovative approach for enhancing the vertical resolution of seismic images, augmenting conventional signal processing with deep learning algorithms and modeling techniques. The algorithm outlined in this study is specifically designed to be applicable across diverse datasets, operating independently of well data. It effectively addresses noise reduction, distinguishing geological features from disturbances and by improving the signal-to-noise ratio, and implements enhancements to the frequency spectrum through a nonlinear approach. The research endeavors to create an unconventional methodology, focusing on vertical resolution improvement. Notably, the effectiveness of this approach is demonstrated using real seismic data collected from the South Caspian Basin. Detailed results are presented, underscoring the potential to significantly enhance the accuracy and reliability of seismic exploration procedures in the realm of geophysics.