基于非抽样复轮廓波变换的图像去噪算法研究

Study of image denoising based on nonsubsampled complex contourlet transform

将二维双树复小波变换(DT-CWT)与非抽样方向滤波器组(NSDFB)相结合,构造一种新的非抽样复轮廓波变换(NSCCT),并对其平移不变性作了相应证明。同时利用对称的正态逆高斯(NIG)分布先验概率模型和贝叶斯最小均方算法,提出一种基于NSCCT的图像去噪算法。实验结果表明,本文构造的NSCCT能够有效地抑制伪Gibbs现象,并且具有更丰富的方向分量,因而在图像的细节和纹理表现方面有一定的优势。

A novel nonsubsampled complex contourlet transform(NSCCT)is constructed by combining the two dimensional dual-tree complex wavelet transform(DT-CWT)and nonsubsampled directional filter bank(NSDFB).And an algorithm of image denoising based on nonsubsampled complex contourlet transform is proposed by using the symmetric normal inverse Gaussian distribution prior and a Bayesian minimum mean squared error estimator.The experimental results show that the nonsubsampled complex contourlet transform can restrain Gi...