Learning Event Guided High Dynamic Range Video Reconstruction

a) LDR images (b) Events (c) E2VID [60] (d) Han et al. [24] (e) Ours Figure 1. Given hybrid inputs of (a) LDR video and (b) stacked events, the HDR video can be reconstructed using different methods shown in (c) E2VID [60], (d) Han et al. [24], and (e) the proposed HDRev-Net. The samples here are tested on synthetic data (top row) and real data (bottom row) respectively. The proposed method is able to generate the HDR video with more details and less flickering effects.