Picture quality evaluation methods could be utilized to assess image algorithms′ performance.The traditional objective methods cannot necessarily represent the perception of error

and subjective methods such as Mean Opinion Score（MOS）

extremely depends on observer′s experiences and motivations.Perceptual Mean Square Error（PMSE） is defined on the basis of Human Visual System（HVS） and then a model is developed to explain this definition.Three components such as Amplitude Nonlinearity

Modulation Transfer Function（MTF）

Orientation and Scale Selectivity contribute to the output perceptual error.In them

B Spline shiftable and Multiscale Wavelet Transform is used to match eye′s Orientation and Scalability.An error segmentation process decomposes perceptual error to four sorts of errors which are then calculated by Mean Square Measurement.PMSE is thus the weighted sum of several principal components after the Principal Component Analysis.The model is tested on images finally

and the results show that PMSE correlates well with MOS.