一种应用于异质扩散的噪声腐蚀新算子

提出了一种新的基于偏微分方程的形态学腐蚀算子，该算子能够直接对梯度图像进行噪声抑制和边缘增强，在异质扩散系数的求取过程中，摈弃了传统的在低分辨率条件下计算图像梯度的方法，而是直接利用该算子对梯度图像进行噪声抑制和边缘增强．与基于高斯光滑以及传统的形态学预滤波方法相比，新的扩散系数具有更好的边缘定位能力和对噪声的鲁棒性．实验结果表明基于该算子的异质扩散滤波新方法具有更好的图像光滑和细节保持性能．

The application of noise erosion operator to anisotropic diffusion

This paper proposed a morphological erosion operator based on partial differential equation (PDE). Its excellent performances on gradient image, noise removal and edge preserving were demonstrated. Based on this operator, an evaluating scheme of diffusion coefficients was introduced: the noise and edge of gradient image was controlled and preserved directly by this morphological erosion operator rather than by pre-smoothing input image with Gaussian function or the conventional morphological operator. The new diffusion coefficient had robustness to noise and better localization of edges. Experiments demonstrated that this anisotropic diffusion scheme could efficiently reduce image noise and sharpen the object boundaries.