Low storage and efficient 3D reverse time migration method based on eikonal equation

Abstract In 3D reverse time migration (RTM), high computational costs and massive storage requirements remain major challenges that are largely determined by the different imaging conditions. The cross-correlation imaging condition (CCIC) provides high imaging resolution but requires storing a large volume of source-wavefield data on the disk and subsequently reading these data back into the memory during imaging. Consequently, the substantial storage demand, intensive input/output operations, and limited data transfer bandwidth substantially reduce computational efficiency. In contrast, the excitation amplitude imaging condition (EAIC) alleviates these limitations by determining the imaging time through the identification of the maximum amplitude of the source wavefield at every grid. However, peak amplitude selection becomes unstable in complex media, producing discontinuous imaging times and distorted excitation amplitudes, leading to reduced imaging accuracy and poor noise robustness. To address this challenge, we propose a novel method for computing the imaging time by solving the 3D eikonal equation using an adaptive finite-difference scheme that adaptively selects appropriate local operators according to the dominant wave propagation modes, thereby providing more accurate traveltimes for imaging time estimation. The proposed method retains the low storage requirement and high computational efficiency of the EAIC while significantly enhancing the resolution of the migration images. Numerical experiments on 3D simple and complex Salt models demonstrate that our proposed method achieves imaging accuracy comparable to that of CCIC while avoiding the computational bottlenecks associated with massive wavefield storage and data transfer. Moreover, it effectively overcomes the limited imaging accuracy and poor noise robustness of the conventional EAIC, thereby showing considerable potential for practical applications.