As traditional pixel-difference statistics and structural similarity (SSIM) cannot well reflect subjective evaluation results，a novel image quality assessment method based on local information distortion modeling is proposed.According to some characteristics of human visual system (HVS)，models are established for the pixel gray distortion，local contrast distortion and local structure distortion.Taking local variance as weight，the final image quality assessment metric is obtained.The experiments on the LIVE database indicate that the proposed method has high prediction accuracy and consistency on distortion types like jp2k，jpeg，gblur and fastfading，and it is better than the SSIM method.Also，a fairly good prediction result is achieved for wn distortion type.Compared with other assessment methods，the proposed method has good predictive performance，and its calculation is quite simple，and it is more consistent with subjective evaluation.