All-in-Focus Seeing Through Occlusions with Event and Frame

Lixuan Wei, Yichen Liu, Kejing Xia, Lei Yu

Abstract: Event-based Synthetic Aperture Imaging (SAI) methods [1, 3, 4] have demonstrated significant potential for see-through imaging behind dense occlusions, utilizing the low latency and unique capture mechanisms of event cameras. By increasing the effective aperture, these methods blur out-of-focus foreground occlusions and reconstruct occluded targets within the depth of field, effectively addressing challenges posed by extremely dense occlusions. However, as occlusions become sparse, the performance of event-only reconstruction declines due to reduced event data [5]. To mitigate this issue, fusion-based SAI [5] combines event streams and image frames, achieving see-through effects across varying occlusion densities and expanding the technique’s applicability.

External IDs:dblp:conf/prcv/WeiLXY25