中文摘要：

提出一种基于贝叶斯估计和剪切波变换相结合的图像去噪算法。对含有加性高斯白噪声的图像进行剪切波变换,得到各尺度各方向上的剪切波系数;利用剪切波系数的相关性,以当前子带剪切波系数为中心,选取尺寸合适的邻域窗口,以该窗口为单位,在贝叶斯最大后验概率准则下推导出基于剪切波系数为拉普拉斯先验分布的最大后验估计表达式和子带阈值, 再通过软化处理达到系数收缩的目的;对处理后的剪切波系数反变换,得到处理后的图像。实验表明,与传统的小波域去噪算法相比,该方法获得了明显的峰值信噪比增益,主观视觉效果也得到了改善。

英文摘要：

Based on the combination of Bayesian estimation and Shearlet transform, an algorithm for image denoising is proposed. In order to get Shearlet coefficients in all scales and directions, image with additive white Gaussian noise is processed by Shearlet transform. With the dependencies of the Shearlet coefficients, the new method select a proper neighbo-ring window by centering the current coefficient within it, and conclude the MAP expression and sub-band thresholds under Bayesian maximum posterior probability criteria when Shearlet coefficients is thought to have Laplace prior distribution, then made shrinkage on it by soft threshold. Inverse Shearlet transform is performed to the processed coefficients and get the denoised image. The experimental results demonstrate that compared with traditional wavelet domain denoising algorithms, the proposed method not only improves peak ratio of signal to noise but also have a remarkable visual effects.