Region-aware Half Wavelet Attention Network for Mixed Exposure Correction

Images captured under different exposure conditions may contain both over- and under-exposures. Existing methods mainly focus on addressing over- or under-exposure problem in images. In this paper, we introduce an effective Region-aware Half Wavelet Attention Exposure Correction Network (RHECNet). Specifically, we introduce a region-based feature extraction module that effectively converts the region features in the mixed exposure scene into feature space, which is invariant to exposure. Besides, as incorrect exposure inevitably reduces structure information, we introduce an attention restoration block based on wavelet domain features. It combines exposure-invariant features with those extracted from the wavelet domain to restore structural information. RHECNet has competitive performance results on two image exposure Correction datasets in terms of quantitative metrics and visual quality.