基于最小生成树的加权中值滤波算法

Weighted Median Filtering Algorithm Based on Minimum Spanning Tree

根据图像纹理分布特点,提出一种基于最小生成树的加权中值滤波算法。依据最小生成树计算像素点的相关度,对像素点进行初次分类。对初次分类中不能确定性质的像素点,采用模糊理论进行二次分类。根据像素点的分类结果,保持像素点的原有灰度值或采用不同的滤波方法进行滤波处理。仿真实验结果表明,在去除噪声和图像细节保持方面,该算法优于其他中值滤波算法。

For the characteristics of image texture,a weighted median filtering algorithm based on minimum spanning tree is proposed.It calculates pixel correlations according to minimum spanning tree,based on which it classifies pixels for the first time.After that,for the pixels that can not be classified,fuzzy theory is used to classify pixels again in order to detect characteristic of pixels.Different filters are used or original pixels are just preserved. Simulation results show this algorithm has much better effects than others in terms of de-noising and retention of image details areas.