A new image quality assessment approach based on quality parameter variance on image multiple resolution representation is proposed to simulate the human vision system. Firstly

a district wavelet transformation is presented to obtain multiple resolution image representation. Image quality parameter is calculated on each level resolution image. Secondly

the image quality parameter variances are calculated and the remaining error of quality parameter after linear simulation is adopted to represent the smoothness of image quality variance. Finally

the image quality assessment model based on the remaining error of different resolution image representations is proposed. Applying the traditional image quality assessment method and the new model to images of different compression ratio show that the new image quality assessment model can obtain the accordant quality assessment effect both the absolute quality parameter and subjective vision. The new model is fit for human vision.