A new image quality metric based on human visual system

In this paper we propose a new image quality metric based on human visual system (HVS) called PSNR-JND. This metric is based on the values of DCT coefficients of 8×8 pixel block of an image. In particular, it assumes the same formula as the PSNR (Peak Signal-to-Noise Ratio), but it computes the MSE (Mean Squared Error) on the perceptual error calculated by dividing the DCT error by the minimum DCT coefficient able to produce a visible signal called also JND (Just Noticeable Difference) or differential threshold. The differential thresholds were experimentally calculated in order to maximise the correlation between the evaluations of the metric and the human visual system. The metric PSNR-JND has outperformed other reference based quality metrics and demonstrated high correlation with the results of subjective experiments of the database TID2008.