在图像降噪中,小波阈值法可以有效的降低图像的噪声,但阈值选取往往是固定的。而经过小波变换后图像各子带的统计特性有所不同,因此δ的值应根据各子带的特性来选取,即δ应为自适应阈值。另外用阈值法消噪后,为消除在一些不连续点会出现伪吉布斯现象,考虑阈值消噪前对图像进行循环平移,形成自适应阈值与循环平移的有机结合。通过仿真实验证明了小波阈值法比一般的图像消噪方法有所改进。特别是图像的均方误差有很大的降低,同时提高了信噪比。
Threshold method could effectively remove the niose in image denoising. But in traditional practice the hreshold is always fixed. The characteristic of every image subdomain is different after the wavelet transforms, so the value of δ should accord to the subdomains. That is to say δ is an adoptive value. And in such cases as discontinuities points, the Gibbs phenomena will appear. In order to avoid this phenomena, we first apply the method termed "cycle spinning" based on threshold wavelet and then use the tow methods to remove the image noise. According to the results of experiment, the proposed algorithm is much better than the condition deniosing methods. The RMSE of signal has been decreased substantially while the SNR has been increased.