基于曲面特性的图像非线性扩散滤波

利用扩散滤波进行图像降噪的过程中，一个核心问题是，如何控制扩散系数，使得模型在图像信息位置停止扩散，而在噪声处有效地扩散。为了更好地解决此问题，本文采用了一种新的思想，把图像看作是三维空间的一个曲面，这样可以得到图像曲面的两个基本特性:高斯曲率和平均曲率。为了能够在图像进行扩散滤波处理中有效地利用图像在三维空间中的这些曲面特性，文章分析了已有的基于平均曲率或高斯曲率的非线性扩散滤波模型，总结了平均曲率和高斯曲率的特点，并在此基础上，提出了基于混合曲率的扩散滤波模型；该模型作为一种新的基于曲面特性的图像扩散滤波模型，同时利用了图像的高斯曲率和平均曲率，恰当地融合了两种曲率的特点，能够以相对较快的速度滤除噪声，同时保持图像的细节特征。 

Surface Properties Based Image Nonlinear Diffusion Filter

The key difficulty in image denoising using diffusion filter is how to control the diffusion function in different location where the ratio of image information and noise. Aiming to this difficult, this article uses a new idea in which by regarding the intensity images as two-dimensional surface in a three-dimensional space, one can get the two base properties of image surface in the three-dimen-sional space: gauss curvature and mean curvature. Nonlinear diffusion filter using surface properties for noise removal is a new type of noise removal filter. This paper proposes a blend curvature driven diffusion model which synthesizes image surface's Gauss curvature and means curvature information. The main advantage of this model is that it preserves important image structures and has a suitable dif-fusion speed.