关键词: 图像去噪, 抗混叠轮廓波变换, 互信息量

Abstract: For the limitations of image information representation of complex wavelet transform and neighborhood correlation description depiction by NeighLevel，an improved image denoising method is presented.First of all，characters of the direction number selected freely for non-aliasing of non-aliasing contourlet are used，the image edge details are expressed，and the deficiency of complex wavelet for direction information is overcome；then the expression ability is improved by neighborhood information through the mutual information of neighborhood wavelet coefficients.Theoretical analysis and experimental results show that its peak signal to noise ratio increases by 0.6%~7% when the variance of image noise is between 30 and 60 compared to the method named CWT-NeighLevel，meanwhile a better visual result is maintained in the aspect of the edge feature.

Key words: image denoising, non-aliasing contourlet transform, mutual information