双树复小波域的邻域自适应贝叶斯收缩去噪

Neighboring adaptive BayesShrink image denoising in dual-tree complex wavelet transform

为了更加高效去除图像采集或传输中引入的噪声,提出了一种基于双树复小波域的邻域自适应贝叶斯收缩的图像去噪方法,利用了双树复小波变换的平移不变性和更多的方向选择性的优点,并考虑了系数间的局部自适应邻域相关性,以尺度适合的窗口为单位估计相应系数的方差,利用滑窗求其平均作为整个子带的图像方差,通过贝叶斯收缩来处理小波系数,从而实现高效的图像去噪。实验结果证明,该方法取得了很高的峰值信噪比和更好的视觉效果,去噪性能优良。

In order to remove noise which is introduced by image acquisition or transmission more effectively, a neighboring adaptive Bayesian shrinkage image denoising method in dual-tree complex wavelet domain is proposed. This method makes use of the translation invariance and the advantage of more direction selective of the dual-tree complex wavelet transform, and the local adaptive neighborhood correlation of the coefficient is also considered. The variance of the corresponding coefficient of the appropriate neighborhood full inch window is estimated, the average of the variance which is used as the variance of the whole sub-band image is calculated using the sliding window. BayesShrink method is used to handle the wavelet coefficients to achieve efficient image denoising. The experimental results show that the proposed method gets higher PSNR and better visual expression. The denoising performance is excellent.