基于Contourlet变换尺度间相关的图像去噪

Image denoise based on inter-scale correlations of Contourlet transform

Contourlet域数据分析表明,信号的变换域系数在尺度间相关性高,而白噪声则呈弱相关或不相关。通过相关性强弱区分噪声与信号系数,并结合阈值函数,提出了基于Contourlet变换尺度间相关的图像去噪新算法。实验结果表明,新方法去噪后的图像比小波相关去噪算法的PSNR值更高,视觉效果更好,尤其适用于纹理轮廓丰富的图像去噪。

Analyses of Contourlet coefficients indicate that inter-scale contourlet coefficients with respect to signals are highly correlated while the correlations associated with white noise are less or even do not exist. According to the property, an image denoising algorithm based on inter-scale correlations of Contourlet coefficients is proposed combined with threshold functions. Experimental results show that the proposed method outperforms the corresponding wavelet method in terms of both peak-signal-to-noise （PSNR） and visual quality, and it is especially adequate to the images with much texture.