基于PDE's的图像平滑方法

Image Smoothing Method Based on PDE's

传统的图像平滑方法在去除噪声的同时往往会破坏边缘、线条、纹理等图像特征,而基于偏微分方程(PDE's)的各向异性扩散算法则在抑制噪声的同时能够保持这些特征。本文在Perona＆Malik模型基础上引入梯度阈值和高斯平滑核,实验结果表明改进后的平滑方法既能更有效消除孤立噪声点,又可以更好地保持边缘。

The conventional image smoothing methods can eliminate the noise as well as smear the image features such as edges,lines and textures,while the anisotropic diffusion algorithm based on partial differential equations(PDE's) can not only suppress the noise but also preserve these features.An improved method is proposed by adding a gradient threshold and a Gaussian smoothing kernel to the Perona & Malik model,and it's proved by experiments to be able to remove the isolated noise more effectively as well as to preserve the edges better.