Abstract: Dynamic time warping (DTW) can be used to improve the alignment of seismic events and correct waveform distortions in common reflection point (CRP) gathers caused by velocity errors and stretch effects. However, its trace-by-trace processing approach is susceptible to noise interference, which can lead to waveform mismatches and degrade the quality of stacked sections. This paper proposes an improved DTW algorithm incorporating multidimensional constraints. This method processes common-offset gathers (COG) as units and introduces regularization constraints along the time, crossline, and inline directions, effectively suppressing abnormal variations in the time-shift field and enhancing its global consistency and stability. Practical application results demonstrate that the improved DTW algorithm effectively avoids the mismatches associated with conventional DTW, improves the coherence of events in prestack gathers, and enhances the continuity of the stacked section. The proposed method exhibits strong robustness in complex structures and under low signal-to-noise ratio conditions, while offering high computational efficiency, making it suitable for large-scale prestack gather processing. It provides a reliable technique for optimizing prestack gathers in geologically complex areas.