Luminance decomposition and reconstruction for high dynamic range Video Quality Assessment

Jifan Yang, Zhongyuan Wang, Baojin Huang, Jiaxin Ai, Yuhong Yang, Jing Xiao, Zixiang Xiong

Abstract: Highlights•This paper proposes an NR VQA model based on luminance decomposition and reconstruction, which uses the decomposition operation to compress the high dynamic range while effectively preserving the local spatial feature information.•In the reorganisation phase, a residual aggregation module is proposed. Hierarchical features from three regions are aggregated using residuals, which helps to reduce redundancy and represent static spatial features.•This paper presents a simple yet effective dynamic motion-aware feature extractor to capture distortion features in the spatio-temporal domain. The interaction between local motion and spatial distortion is further constructed along the flow of frames.