Abstract:In this paper, we propose a pipeline based on two-dimensional(2D) phase correlation to tackle the challenge of stereo event-based depth estimation under sparse data conditions. Initially, a 2D Fourier transform is applied to image blocks (patches) at corresponding positions of each event on the stereo event frame to extract spectra, thereby capturing the structural characteristics and edge details of events in the frequency domain. Subsequently, the cross-phase spectrum between patches is computed using these spectra. Finally, a trigonometric function expression is derived through an in-verse Fourier transform, where the peak coordinate signifies the disparity of the patch. The proposed method was ex-perimented on the MVSEC dataset, the RPG dataset, and achieved a mean depth error of 0.57m on frames 140 to 1200 in the indoorflying1 sequence, outperforming peer works cited in this paper while utilizing 50% of their events.