A new way of pooling: starting from an image quality measure

This paper investigates a very simple image quality measure, which is defined by a ratio based mathematical formulation. It offers an alternative approach to pooling in the design of human vision based image/video picture quality metric. Other than the popular error sensitivity based pooling technique, this ratio sensitivity based quantitative measure is linearly correlated with the mean opinion score (MOS) and avoids the nonlinear fitting error to a certain extent An evaluation based on image quality testing data from a public source demonstrates that this measure is highly correlated with the MOS. This ratio sensitivity based approach to pooling may be used to improve current perceptual image/video quality metrics.