Low-Complexity Deep No-Reference Light Field Image Quality Assessment with Discriminative EPI Patches Focused
Ping Zhao, Xiaoming Chen, Yuk Ying Chung, Haisheng Li · 2021
Light field image quality assessment (LF-IQA) has attracted increasing research interests due to the fast-growing demands for immersive media experience. In this paper, we propose a low-complexity deep residual convolutional neural network (CNN) based no-reference (NR) LF-IQA approach for potential applications of light field images on consumer electronic devices. For efficient and accurate LF-IQA, discriminative Epipolar Plane Image (EPI) patches are taken as input to our CNN model that employs a dedicatedly designed multitask learning mechanism. The experimental results show that the LF-IQA metric resulted from our approach outperforms state-of-the-art metrics against representative benchmark datasets.