利用非抽样Shearlet域GSM模型进行图像去噪

为有效去除含噪图像中的噪声,提出了一种基于非抽样剪切波域高斯比例混合模型的图像去噪方法。首先建立含噪图像非抽样剪切波系数的局部高斯比例混合模型,然后应用贝叶斯最小二乘法对无噪图像的非抽样剪切波系数进行估计,最后通过非抽样剪切波逆变换得到去噪后的图像。该方法充分利用了非抽样剪切波变换的平移不变性、对图像边缘纹理等细节的高效表示能力以及高斯比例混合模型对非抽样剪切波变换系数局部相关性的概括能力。实验结果表明,与基于小波域高斯比例混合模型的图像去噪方法、曲波域多变量阈值去噪方法以及非抽样剪切波域的硬阈值法相比,该方法不仅能更有效地去除含噪图像中的噪声,提高其信噪比以及与原始无噪图像的平均结构相似度,...

Image Denoising Based on the NSST Domain GSM Model

An image denoising method based on the non-subsampled Shearlet domain Gaussi- an scale mixture model is presented.First,a Gaussian scale mixture model is used to model the correlation of the locally non-subsampled Shearlet coefficients of the noisy image.Then,the noise-free coefficients are estimated by the Bayes least square estimator.Finally,the inverse non-subsampled shearlet transform(NSST) is applied to these estimated Shearlet coefficients to obtain the denoised image.Experimental results show that the proposed method can remove Gaussian white noise while effectively preserving edges and texture information.At the same time,it can achieve a higher PSNR and mean structural similarity than the wavelet based GSM method,the curvelet domain multivariate shrinkage method and the non-subsampled Shearlet domain hard thresholding method.