中文摘要：

针对标准中值滤波算法边缘保持能力较差的缺点，设计了K均值中值滤波算法和递归K均值中值滤波算法。使用K均值方法将中心像素点的邻域数据序列分成两类，将该像素点所属的这类数据的中值作为输出。为了加快算法的速度，提出了两个阈值死和％，保证在不影响结果精确性的同时尽量减少处理时间。传统的K均值算法耗时较高，论文的另一个改进是对K均值的优化处理，大大缩短了算法的处理时间。实验证明改进算法具有较好的噪声抑制能力和边缘保持能力。

英文摘要：

This paper designs a K-means median filter algorithm and recursive K-means median filter algorithm to solve the defects of standard median filter.At first,the sample composed by the current pixel and its neighbors is divided into two classes, and the median of the class which includes the current pixel is computed as the result.To ensure the accuracy and speed up the process,two parameters TL and TH are designed.The standard K-means algorithm is a time-consuming procedure,so the optimization is processed in this paper.The experiment proves that the new algorithms can get better noise reduction effect.