论文部分内容阅读
利用小波变换去噪时小波系数方差的估计对去噪结果影响很大。自然图像小波分解后得到的系数在不同的分辨率中差异很大,所以利用邻域估计中心点方差时,不同分辨率应有不同大小的邻域。首先对在邻域中利用极大似然准则估计中心点方差进行分析,再结合自然图像小波分解后的系数在不同分辨率子带中,根据平稳性和重要性选择邻域的大小。最后进行去噪实验,并取得正交小波分解下理想的去噪性能。
The use of wavelet transform denoising wavelet coefficient estimation of variance has a great impact on the denoising results. The coefficients obtained from the wavelet decomposition of natural images vary greatly in different resolutions. Therefore, when using the neighborhood to estimate the variance of the center point, different resolutions should have different sizes of neighborhoods. Firstly, the variance of the central point is estimated by using the maximum likelihood criterion in the neighborhood, and then the size of the neighborhood is selected according to the stationarity and importance in different resolution subbands by combining the coefficients of the natural image wavelet decomposition. Finally, the denoising experiment is carried out and the ideal denoising performance under orthogonal wavelet decomposition is obtained.