Design of an image distortion measure using spatial/spatial-frequency analysis

It is generally acknowledged that the image quality measures based or inspired by some findings on the human visual system (HVS) mechanisms are more appealing than classical measures such as the PSNR or MSE. Many HVS-based image quality assessment methods have been previously proposed. However, because of the complexity and the influence of the great number of observation parameters on the HVS-based image quality metrics there is no standard measure yet. In this paper, the image distortion metric proposed in A. Beghdadi et al. (2003) is reviewed and some improvements are suggested. This metric makes use of the joint spatial/spatial-frequency representation and applies a relatively simple nonlinear energy detection procedure. This simple distortion measure does not take into account any well-established model for the masking effect. The consistency of the proposed measure with subjective human evaluation is demonstrated on some standard image distortions.