Joint RGB-Spectral Decomposition Model Guided Image Enhancement in Mobile Photography

• DOI: 10.48550/arxiv.2407.17996
• Publication Date: 2024-07-25
• Authors: Kailai Zhou; Lijing Cai; Yibo Wang; Mengya Zhang; Bihan Wen; Qiu Shen; Xun Cao
• Keywords: RGB color model; Computational photography

The integration of miniaturized spectrometers into mobile devices offers new avenues for image quality enhancement and facilitates novel downstream tasks. However, the broader application spectral sensors in photography is hindered by inherent complexity images constraints imaging capabilities. To overcome these challenges, we propose a joint RGB-Spectral decomposition model guided framework, which consists two steps: prior-guided enhancement. Firstly, leverage complementarity between RGB Low-resolution Multi-Spectral Images (Lr-MSI) to predict shading, reflectance, material semantic priors. Subsequently, priors are seamlessly integrated established HDRNet promote dynamic range enhancement, color mapping, grid expert learning, respectively. Additionally, construct high-quality Mobile-Spec dataset support our research, experiments validate effectiveness Lr-MSI tone task. This work aims establish solid foundation advancing vision photography. code available at \url{https://github.com/CalayZhou/JDM-HDRNet}.