基于非下采样Contourlet变换和谱图理论的扩散去噪

提出了一种基于谱图理论的热扩散方程图像去噪方法.该方法用非下采样的Contourlet变换提取图像的边缘和轮廓等几何特征,并将提取的特征用来构造图的权重函数,将扩散方程建立在图上,用热核和拉普拉斯矩阵实现图像的去噪.仿真实验结果表明,该方法能够有效去除高斯噪声,较完整地保持图像中的边缘等细节信息,在去噪性能上优于其他的偏微分方程去噪方法.

Image diffusion denoising based on spectral graph theory and nonsubsampled Contourlet transform

A diffusion method based on spectral graph theory for image denoising is proposed, which uses nonsubsampled contourlet transform to capture the geometric feature of the image. After that, a new graph weighting function is constructed based on the captured geometric feature. The heat diffusion equation is generated on a graph. Meantime, a heat kernel and Lapician matrix are used to filter the noisy image. Simulation experiments and comparisons with standard images illustrate the effectiveness of the method. Compared with some other existing methods, the proposed method can effectively reduce Gaussian noise and preserve image edges. Its performance is superior to other partial differential equation methods.