RCNet: Deep Recurrent Collaborative Network for Multi-View Low-Light Image Enhancement

Abstract

Scene observation from multiple perspectives brings a more comprehensive visual experience. However, acquiring multiple views in the dark causes highly correlated views alienated, making it challenging to improve scene understanding with auxiliary views. Recent single image-based enhancement methods may not provide consistently desirable restoration performance for all views due to ignoring potential feature correspondence among views. To alleviate this issue, we make the first attempt to investigate multi-view low-light image enhancement. First, we construct a new dataset called Multi-View Low-light Triplets (MVLT), including 1,860 pairs of triple images with large illumination ranges and wide noise distribution. Each triplet is equipped with three viewpoints towards the same scene. Second, we propose a multi-view enhancement framework based on the Recurrent Collaborative Network (RCNet). To benefit from similar texture correspondence across views, we design the recurrent feature enhancement, alignment, and fusion (ReEAF) module, where intra-view feature enhancement (Intra-view EN) followed by inter-view feature alignment and fusion (Inter-view AF) is performed to model intra-view and inter-view feature propagation via multi-view collaboration. Additionally, two modules from enhancement to alignment (E2A) and alignment to enhancement (A2E) are developed to enable interactions between Intra-view EN and Inter-view AF, utilizing attentive feature weighting and sampling for enhancement and alignment. Experimental results demonstrate our RCNet significantly outperforms other state-of-the-art methods.