Seismic Imaging: An Overview and Parallel Implementation of Poststack Depth Migration

TL;DR

This paper reviews seismic imaging with a focus on parallelizing poststack depth migration using CUDA, highlighting the potential for parallelism at each depth level despite data dependencies, and evaluates its performance.

Contribution

It introduces a CUDA-based parallel implementation of poststack depth migration and assesses its performance compared to sequential algorithms.

Findings

Parallelization is feasible at each depth level despite data dependencies.

CUDA implementation accelerates seismic imaging processing.

Performance improves with increasing problem size.

Abstract

Seismic migration is the core step of seismic data processing which is important for oil exploration. Poststack depth migration in frequency-space (f-x) domain is one of commonly used algorithms. The wave-equation solution can be approximated as FIR filtering process to extrapolate the raw data and extract the subsurface image. Because of its computational complexity, its parallel implementation is encouraged. For calculating the next depth level, previous depth level is required. So, this part cannot be parallelized because of data dependence. But at each depth level there is plenty of roam for parallelism and can be parallelized. In case of CUDA programming, each thread calculate a single pixel on the next depth plan. After calculating the next depth plan, we can calculate the depth row by summing over all the frequencies and calculating all the depth rows results in the final migrated image. The poststack depth migration is implemented in CUDA and its performance is evaluated with the sequential code with different problem sizes.