A ratio sensitive image quality metric
Zhizhong Zhe, Hong Ren Wu · 2005
Many applications require a fast quality measure for digitally coded images, such as the mobile multimedia communication. Although the widely used peak signal-noise ratio (PSNR) has low computation cost, it fails to predict structured errors that dominate in digital images. On the other hand, human visual system (HVS) based metrics can improve the prediction accuracy. However, they are computationally complex and time consuming. This paper proposes a very simple image quality measure defined by a ratio based mathematical formulation that attempts to simulate the scaling of the human visual sensation to brightness. The current results demonstrate that the novel image quality metric predicts the subjective ratings better and has lower computation cost than the PSNR. As an alternative to the PSNR, the proposed metric can be used for on-line or real-time picture quality assessment applications.