一种改进的方向扩散方程滤波方法

An improved directed diffusion equation filtering method

利用方向扩散方程去噪时,噪声滤除的同时边缘也很快模糊了。基于这一缺陷,本文用各向异性扩散算子代替了方向扩散方程第一项中的拉普拉斯算子,并在方程的两个扩散项前加上了不同的扩散系数,以保证算法既能快速扩散去除噪声,又能较好保留边缘。而且,当新模型中的初始逼近图像退化为常数时,本文模型就退化为Perona-Malik扩散方程,因此Perona-Malik扩散方程是本文新模型的一个特例。实验结果和客观数据分析均表明本文算法在保留边缘方面具有明显的效果。

When denoising with the method of directed diffusion equation,the edges will be blurred quickly. To make up for this disadvantage, the Laplace operator in the directed diffusion equation is replaced by the anisotropic diffusion operator, diffusion coefficients are added in the front of the two diffusion items. It is ensured that the noise can be removed quickly and the edge can be preserved better. When initial approximation image degenerates to constant, the new model will degenerate to Perona-Malik diffusion equation. In some sense, Perona-Malik diffusion equation is a special case of the new model. Experiment results and objective data all indicate the efficiency of our new model.