Multi-exposure image fusion via perception enhanced structural patch decomposition

Abstract

Multi-exposure image fusion (MEF) is an affordable and convenient option for high-dynamic-range imaging. Current MEF methods are prone to visually unrealistic results since they take no account of perceptual factors. To address this problem, a multi-exposure image fusion method is proposed based on perception enhanced structural patch decomposition, namely PESPD-MEF. An image patch is first decomposed into four components: perceptual gain, signal strength, signal structure, and mean intensity. Then, the enhancement rule is designed for perceptual gain, and the latter three elements are fused independently in different ways. Finally, the fused components are aggregated to generate informative and perception-realistic results. Moreover, the multi-scale framework is adopted to boost the fused performance. The proposed method is also extended to address single low-light image enhancement issue. Experimental results demonstrate that the proposed approach outperforms the state-of-the-art methods by a large margin in terms of perceptual realism. The source code is available at https://github.com/Junchao2018/PESPD-MEF.