基于DCT及高频屏蔽窗的小波图像去噪方法

Wavelet image denoising based on DCT and high-frequency shielding windows

为精确估计噪声方差,提出一种基于离散余弦变换及高频屏蔽窗的去噪方法,利用离散余弦变换(DCT)对小波高频子带进行局部特征提取,在此基础上对由噪声引起的小波系数的幅值进行估计,从原小波系数中屏蔽噪声部分的幅值,并用小波逆变换对剔除噪声分量后的小波系数进行恢复,从而得到去噪后的图像.该方法避开了噪声方差的估计,并在去噪效果和保留细节方面要好于传统的阈值去噪方法.

In order to estimate of noise variance accurately, this paper proposes a new kind of method which is based on discrete cosine transform （DCT） and high-frequency shielding windows. First, discrete cosine transform is used to extract local features; and then, it is estimated that the amplitude of the wavelet coefficients which is caused by noise; finally, noise amplitude is subtracted from the original wavelet coefficients and the inverse wavelet transform is performed to obtain the denoised image. The method avoids the noise variance estimate, and the effect of denoising and detail of retention are better than traditional methods.