Multiple Levels Perceptual Noise Backed Visual Information Fidelity for Picture Quality Assessment
Chenchen Peng, Mixia Wu, Kaiyuan Liu · 2022
Image Quality Assessment (IQA) is a crucial aspect of image processing. A great deal of works have improved the performance of IQA algorithms, however, the existing classical Visual Information Fidelity (VIF) algorithm has minor limitations: at low scales, fidelity information of images is under-extracted; the variation of perceptual noise of Human Visual System (HVS) with scale is not fully taken into account. In response, this paper proposes an improved VIF-Multiple Levels Perceptual Noise Backed Visual Information Fidelity (MPNVIF), which optimizes the internal algorithm of image information extraction and utilizes multiple levels perceptual noise to enhance the evaluation performance of VIF. Finally, we perform a comparison experiment of MPNVIF algorithm, VIF algorithm and some other objective IQA algorithms on the Industrial Scene Image Database (ISID) database. The results show that compared to related algorithms, the MPNVIF proposed in our paper has better performance on the ISID database.