通常，大部分自然图像中同时存在颗粒噪声和高斯噪声，而单纯地用中值滤波算法或均值滤波难以同时尽可能地消除混合噪声。针对这一问题，L ee和 Kassam提出了一种改进的均值滤波算法 Modified Trim med Mean(MTM)，虽然 MTM算法的滤波效果相对于传统的平滑算法已有了很大的改善，但是 MTM的滤噪能力在很大程度上受到了阈值的限制。在分析 MTM算法和传统平滑算法结构特点的基础上提出了一种改进的自适应中值滤波算法。该算法对含有混合噪声的图像上每一点的 N× N区域应用自适应算子。对于不同的图像区域，算子也相应地有所不同，其中算子中的权值选取依赖于区域的灰度中值，且当某点的灰度越接近灰度中值，其权值就相应地越大。实践证明，新算法的处理结果优于传统的滤波算法和 MTM滤波，且没有阈值限制

Usually,there is guass noise and isolated noise in many nature images simultaneously.And it is difficult to getting rid of guass noise and isolated noise only by Median filter or Mean filter at the same time. In allusion to this question, Lee and Kassam proposed an improved algorithm of mean filter. Although they have maked an great improvement, the effect of MTM is still not ideal. It is a threshold that consumedly affect the effect of MTM.In this article ,we preposed an improved method of median.filer on the basis of analysing the character of MTM and traditional filter method. This new method applies auto-adapted operators on theN×Narea of every point in the processed image.For the different area of the image,the operators are different too.The election of operator weight mostly depends on the median ofN×Narea. The more the gray value be close to the median ,the more the weight of operator is strong. In real application, the new method not only shows better performance than the MTM filter, but also is not affected by any threshold.