Learning Depth From Focus With Event Focal Stack

Depth from focus (DfF) estimates depth by determining the moment of maximum focus from multiple shots at different focal distances, that is, the focal stack. However, the limited sampling rate of conventional optical cameras makes it difficult to obtain sufficient focus cues during the focal sweep, leading to the lack of details in depth estimation. Inspired by biological vision, the event camera records intensity changes over time in extremely low latency, which provides more temporal information for focus time acquisition. In this study, we propose the event-based depth from focus (EDFF) network to estimate depth from the event focal stack (EFS). Specifically, we utilize the event voxel grid to encode intensity change information and project event time surface into the depth domain to preserve per-pixel focal distance information. A focal-distance-guided cross-modal (FDCM) attention module is presented to fuse the information mentioned above. In addition, we propose a multilevel depth fusion block (MDFB) designed to integrate results from each level of a U-Net-like architecture and produce the final output. Furthermore, two EFS datasets for depth estimation are built for training and evaluating our network. Extensive experiments validate that our method outperforms existing state-of-the-art approaches.