Abstract: This paper proposes a new concept of the Mean Neighbor（MN） based on the analysis of the popular filtering algorithms, designs and implements the corresponding filtering algorithm. The new method chooses the window size of the filter adaptively, and calculates the mean value after removing the noise. It proposes three ways of calculating the MN and two ways of judging the noise, and combines them to design nine filtering algorithms. The experiments show that the 7th filtering algorithm has got the best filtering performance after comparing the above nine filtering algorithms with the mean filter, the median filter and some other improved filtering algorithms. The 7th filtering algorithm is more pervasive and can deal with different noise types.

Key words: Mean Neighbor（MN）, adaptive filtering, fast filtering, median filtering, mean filtering

摘要： 对常见的滤波算法进行分析，在此基础上提出“近均值”的概念，设计相应的滤波算法。对滤波窗口尺寸采用自适应的方式，在排除部分可疑噪声数据后，对剩余数据计算均值。提出三种计算近均值的规则和两种噪声判断方案，组合设计了九种滤波算法。在与均值滤波、中值滤波及部分改进的滤波算法进行实验对比后，确定提出的第七种滤波算法具有更好的滤波效果，能适应不同类型的噪声，其普适性和实用性进一步增强。

关键词: 近均值, 自适应滤波, 快速滤波, 中值滤波, 均值滤波