Abstract: In light of without regard of the noise statistical distribution, and only using Fast Fourier Transform（FFT） and wavelet transform to image noise reduction bringing an image distortion, This paper proposes a method of image denoising based on the transform domain and noise estimation. According to FFT and wavelet transform on image effective expression of different emphases and image noise with statistical characteristic in different transform domains, the paper presents image Fourier transformation, then constructs transfer function H with noise statistics, finally obtains the first denoising image after using the Wiener filter for noise reduction. And it carrays out denoising image wavelet transform, then uses threshold denoising method and MinimumMean Square Error（MMSE） criterion estimation method to get the second denoised image according to the noise with different characteristics in wavelet of various scales and the same. Simulation experiments show that the algorithm can effectively improve the effect of noise reduction which the denoising images contain less noise with no distortion.

Key words: transform domain, noise estimation, image denoising, MinimumMean Square Error（MMSE） criterion

摘要： 针对不考虑噪声的统计分布，仅使用傅里叶变换或小波变换对图像进行降噪处理会带来图像的失真（扭曲）的问题，提出基于变换域和噪声估计的图像去噪方法。算法根据傅里叶变换和小波变换对图像的有效表示侧重点不同，以及图像噪声在不同变换域下的统计特性，提出先将图像进行傅里叶变换，根据噪声的统计特性构造传递函数H，使用Wiener滤波器进行降噪处理，得到一次降噪图像；再对图像再进行小波变换，根据噪声在小波的各尺度下，以及同一尺度下的不同特性，分别采用软门限降噪法和MMSE准则的降噪方法，得到二次降噪图像。仿真实验证实，该算法能有效提高降噪效果，降噪后的图像不失真，包含噪声少。

关键词: 变换域, 噪声估计, 图像去噪, MMSE准则