基于非抽样小波的多阈值去噪

利用小波系数之间的相关关系，对变换后的小波系数进行分类，提出了基于多阈值的去噪方法。该方法利用非抽样小波变换的冗余性来寻找小波系数之间的依赖关系，把变换后系数分成与图像细节特征相关的系数、平滑区域相关的系数以及噪声相关的系数，对不同的系数采用不同的阈值处理策略以改进去噪效果。实验结果表明，与其它去噪方法相比，该方法不仅较好的保留了图像特征，而且具有良好的视觉效果，峰值信噪比也有较大幅度的提高。

A Multi-threshold Image Denoising Method Based on Undecimated Wavelet

A method based on multi-threshold is proposed, which employs the relationship among the coefficients and classification of the coefficients after the wavelet transformation. The redundance of undecimated wavelet transform is used to find the relationship of the coefficients, which are classified into three groups, the coefficients related to image feature, the coefficients associated to hoise and the coefficients of the smooth area. Different methods are employed to deal with the different groups of coefficients in order to improve the performance of the denoising. Experimental results show that compared with the other algorithms, this algorithm not only preserves image feature, but also obtains better visual effect. Its PSNR is also greatly improved.