Abstract: Reverse time migration based on the two-way wave equation is a high-precision imaging method that handles full wavefields without dip angle limitations, thereby enabling accurate imaging of complex structures. However, the conventional strategy of storing the entire source wavefield requires repeated disk read/write operations at each time step, which not only consumes substantial storage but also introduces significant I/O latency, imposing severe constraints on computational efficiency. To address this issue, we adopt two storage strategies to optimize the conventional imaging algorithm: one involves sampling and storing the source wavefield based on the Nyquist sampling theorem to reduce I/O frequency, and the other stores only the effective boundary wavefield and reconstructs the source wavefield during backpropagation. The systematic analysis of both optimized imaging algorithms demonstrates that both can effectively reduce storage requirements and significantly enhance computational efficiency. Numerical examples show that the imaging accuracy of the two algorithms is basically the same. When sampling and storing the source wavefield, the storage interval needs to be reasonably set to balance imaging requirements and computational efficiency. In contrast, the boundary wavefield reconstruction strategy requires more GPU video memory. When applying GPU acceleration to large 3D models, the imaging algorithm should be reasonably selected.