Multivariate Statistical Analysis of Existing Image Quality Metrics Over Turbulent Images

State-of-the-art objective image quality metrics are summarized and analyzed from a new perspective. Performance comparisons of existing metrics are firstly conducted on simulated turbulence-degraded images. Then exploratory factor analysis method is used to uncover the latent relationships among existing metrics. Experimental results show that existing metrics are highly correlated to each other, and can be separated into four groups, each measuring the qualities from one aspect. Extracting the communalities of them is believed to be helpful in developing a novel metric.