平移不变与系数放大小波变换图像消噪方法

Image denoising method based on translation invarian and coefficient magnification

平移不变量小波变换在图像降噪中的应用,主要通过阈值方法来有效的降低图像的噪声,但它的结果中会出现诸如伪吉布斯现象之类的情况。为消除此类情况,将平移不变量小波变换引入到小波图像降噪中,并结合阈值方法进行消噪处理,同时在阈值处理前对分解后的高频与低频系数进行适当放大,从而形成平移不变量与系数放大法的有机结合。经仿真实验,证明这种方法比一般的图像消噪方法有很大改进,特别是图像的均方误差有很大的降低,提高了信噪比。

In this paper a method for wavelet transform based on translation invariant in the image denoising area is researched. Threshold method could effectively eliminate the noise. However, some cases such as Gibbs phenomena would appear in its result. In order to eliminate this phenomena, a new approach for wavelet transform based on translation invariant is presented in the image denoising. Moreover, the threshold method is combined to eliminate the image noise. Before dealing with threshold, the coefficients of high-frequency and low-frequency are magnified properly after decomposing. Results of experiment show that the given algorithm is much better than the common deniosing methods. The RMSE of signal has been phenomenally decreased while the SNR has been increased.