Abstract: To constrain the drawback of the image de-noising due to the lack of translation invariance and redundancy of original Contourlet transform, a new image de-noising algorithm was proposed based on multi-wavelet nonsubsampled Contourlet transform and Bayes Shrink adaptive threshold, which used multi-wavelet for multi-scale decomposing and nonsubsampled filter banks for multi-direction decomposing, then improved Bayes Shrink adaptive threshold method according to the relation among decomposed sub bands. The experimental results show that the Signal-to-Noise Ratio (SNR) values of the de-noising images are improved significantly compared with the existing algorithm. The proposed algorithm has reduced pseudo-Gibbs phenomena effectively and preserved more details and edge information of the image.

Key words: multi-wavelet nonsubsampled Contourlet transform, image de-noising, adaptive threshold, Bayes Shrink method

摘要： 为抑制Contourlet变换的非平移不变性和冗余性给图像去噪所带来的图像失真等缺陷，提出一种新的基于多小波—非采样Contourlet变换和基于Bayes Shrink的自适应阈值去噪算法：首先利用多小波对图像进行多尺度分解并结合非下采样方向滤波器组进行方向分解，接着根据分解所得到的各方向子带的关系，改进了Bayes Shrink自适应阈值取值方法，对图像进行去噪处理。实验结果表明：该算法去噪后图像的信噪比(SNR)与已有算法相比，有了明显的提高，有效地抑制了原Contourlet变换所造成的伪Gibbs现象，更好地保留了图像的细节信息。

关键词: 多小波—非采样Contourlet变换, 图像去噪, 自适应阈值, Bayes Shrink方法