基于各向异性扩散滤波的边缘提取算法

Edge extraction algorithm based on anisotropic diffusion filtering

针对Canny边缘检测算法在滤波方面的不足,提出了一种基于各向异性扩散滤波的边缘提取算法,用各向异性扩散的平滑方式代替Canny算法中的高斯滤波,弥补Canny算法中滤波方法的不足.研究了各向异性扩散的P-M方程,得出了其离散化形式,并确定了其中分量的表达式,进而用Matlab编程实现了对图像的各向异性扩散处理.通过误差和结构相似度评定,得出经过各向异性扩散处理的图像滤波效果优于高斯滤波.对经过两种滤波方法处理后的图像用Canny算法提取边缘,通过灰度值比较,验证了各向异性扩散处理在边缘特征保持方面优于高斯滤波.

Aiming at the shortcomings of Canny edge detection algorithm in the filtering,an edge extraction algorithm based on anisotropic diffusion filtering was proposed.The Gauss filtering in Canny algorithm was replaced by the anisotropic diffusion smooth approach,which could patch the disadvantages of filtering method in Canny algorithm.The anisotropic diffusion P-M equation was studied,its discretization forms were obtained,and the corresponding component expression was determined.In addition,the anisotropic diffusion treatment of graphs was realized with the Matlab programming.Through the evaluation of both errors and structure similarity,it is found that the graphic filtering effect processed by the anisotropic diffusion is superior to that obtained with the Gauss filtering.The edges of graphs processed by above-mentioned methods were extracted with the Canny algorithm.Through comparing the gray value,it is proved that the anisotropic diffusion treatment is superior to the Gauss filtering in the edge feature-preserving.