基于小波-Contourlet变换的Cycle spinning硬阈值图像去噪方法

综合考虑小波变换的特点以及Contourlet变换对二维光滑图像有很好的近似特性和对曲线有更好的"稀疏"表示的特点,提出了一种基于小波-Contorlet变换的图像硬阈值去噪方法,并用Cycle spinning去除图像中小波-Con- tourlet变换平移变异性而产生的伪吉布斯现象。实验结果表明:该方法与采用Cycle spinning的小波和Contourlet去嗓算法相比,PSNR分别提高了0.4～1.6和0.2～1.0。与其它去噪算法相比,这种方法能有效地去除图像中的噪声,具有更高的PSNR值,能更好地保留图像的纹理和细节。

Wavelet-based Contourlet transform for image de-noising using cycle spinning

Due to the characteristics of the wavelet transform,the good approximation of the Contourlet transform to the 2D smooth image and the ability of the better"sparse"expression of curve,a new scheme for image de-noising based on the wavelet-based contourlet transform is proposed.It uses cycle spinning to dislodge the Gibbs-like phenomenon which is caused by the translation variability of the wavelet transform in the image.Experimental results indicate that compared to the de-noising schemes of cycle spinning-based the wavelet and the contourlet transforms,this scheme's PSNR increase about 0.4～1.6 and 0.2～1.0 resectively.Compared to other de-noising schemes,this scheme can eliminate the noise in images more effectively, which has much larger values of PSNR and tan preserve the detail and the texture of the image more perfectly.