基于上下文模型的混合傅里叶-小波图像降噪方法

Fourier-wavelet image reduction using context-based model

针对小波变换和傅里叶变换去除图像噪声时各具不同的优点和不足,提出一种基于上下文模型的混合傅里叶小波图像降噪方法。首先在傅里叶-域中估计原始图像的功率谱密度,运用维纳滤波器降噪,降低原始图像噪声水平;再在小波域中通过基于上下文模型的自适应阈值法去除剩余噪声;在小波域中使用平稳小波变换分解图像信号得到分解后的系列小波系数,根据小波系数间的相关性,利用上下文模型求取小波系数的方差,将其代入由GGD模型估计出的阈值表达式得到自适应阈值,再用软阈值函数对小波系数进行处理,最后将处理后的小波系数进行小波逆变换完成去噪。仿真结果表明:该方法不仅能够有效滤除图像噪声,而且能够保留图像的边缘细节信号,抑制降噪引起的吉布斯现象。

The strengths and weaknesses of Fourier transform and wavelet transform were discussed in the process of image-denoising.The image noise reduction methods were proposed using Fourier-wavelet transforms context-based model.The original image noise level was reduced by estimation of power spectral density of the original image in the Fourier domain and the use of Wiener filter.The remaining noise was removed by using adaptive wavelet threshold method based on context modeling.A series of wavelet coefficients were obtained by using the stationary wavelet transform to decompose the image signal.According to the correlation among those coefficients,the variance of wavelet coefficients was calculated by using the context model.The adaptive threshold was acquired by substituting the variance into the threshold expression estimated by the GGD model.Processing wavelet coefficients with soft thresholding function,the image denoise was completed by means of the wavelet inverse transform to wavelet coefficients.Simulation results show that the methods proposed can not only effectively filter out image noise,but also keep the edges of the image detail signal well and inhibit Gibbs phenomenon caused by noise reduction.