Premium
Image quality assessment by an efficient correlation‐based metric
Author(s) -
Lin LiHui,
Chen TzongJer
Publication year - 2020
Publication title -
concurrency and computation: practice and experience
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.309
H-Index - 67
eISSN - 1532-0634
pISSN - 1532-0626
DOI - 10.1002/cpe.5794
Subject(s) - lossless compression , lossy compression , metric (unit) , image quality , artificial intelligence , image compression , correlation , computer science , image (mathematics) , human visual system model , consistency (knowledge bases) , quality (philosophy) , computer vision , pattern recognition (psychology) , data compression , computation , image processing , mathematics , algorithm , philosophy , operations management , geometry , epistemology , economics
Summary Image quality can be measured visually. In the human visual system, a compressed image can be judged by the human eye. Image quality may not be perceived to decline in a region with low compression. However, image quality clearly declines in a region with high compression. As image compression increases, image quality gradually transitions from visually lossless to lossy. In this study, we aim to explain this phenomenon. A few images from different datasets were selected and compressed using JJ2000 and Apollo, which are well‐known image compression algorithms. Then, error‐based and correlation‐based metrics were applied to these images. The correlation‐based metrics agree with human‐vision evaluations in experiments, but the error‐based metrics do not. Inspired by the positive result of the correlation‐based metrics, a new metric named the simple correlation factor (SCF) was proposed to explain the aforementioned phenomenon. The results of the SCF show good consistency with human‐vision results for several datasets. In addition, the computation efficiency of the SCF is better than that of the existing correlation‐based metrics.