摘要
提出了一种用各向异性双变量拉普拉斯函数模型去模拟NSCT域的系数的图像去噪算法,这种各向异性双边拉普拉斯模型不仅考虑了NSCT系数相邻尺度间的父子关系,同时满足自然图像不同尺度间NSCT系数方差具有各向异性的特征,基于这种统计模型,文中先推导出了一种各向异性双变量收缩函数的近似形式,然后基于贝叶斯去噪法和局部方差估计将这种新的阈值收缩函数应用于NSCT域,实验结果表明文中提出的方法同小波域BiShrink算法、小波域ProbShrink算法、小波域NeighShrink算法相比,能够有效地去除图像的高斯噪声,提高了图像的峰值信噪比;并较完整地保持了图像的纹理和边缘等细节信息,从而明显改善了图像的视觉效果。
A new method for image denoising based on the Nonsubsampled Contourlet Transform （NSCT） domain with anisotropic bivatiate shrinkage is proposed. This method presents a novel image denoising algorithm based on the modeling of NSCT coefficients with an anisotropic bivariate Laplacian distribution function . The s anisotropic bivariate Laplacian model not only captures the child-parent dependency between NSCT coefficients,but also fits the anisotropic property of the variances of NSCT coefficients in different scales of natural images.With this statistical model,a closed-form anisotropic bivariate shrinkage function in the framework of Bayesian denoising is derived,then the new shrinkage function is exploited in the NSCT domain with local marginal varience estimation . The proposed method can effectively reduce Gauss noise in remote sensing image and improve the image of the peak signal-to-noise ratio;This method utilizes the translation invariant of NSCT transform to hold-up the de-noising pseudo Gibbs effect, and preserves the image texture and edge detail etc informations, thus obviously ameliorate the visual effect of the image.
出处
《电子设计工程》
2012年第18期178-181,共4页
Electronic Design Engineering
