Abstract: This paper proposes a new filtering method.First，we have got detail image from noise image and image of median filtering.Have used a noise detection method to get another image，which detains noises but removes details in detail image.Using this image，we can detect noises in corrupted image more accurately.To every pixel in noise image，the output of filter is the linear combination of original pixel value and the median of pixels inside windows of noise image.If the chance of corrupt pixel being noisy is big，then it will improve much in filtering process.Compared with other methods of median type，the proposed method can not only remove impulse more effectively，but also detain more details.

Key words: impulse noise, median filtering, noise ratio

摘要： 提出了一种新的滤波方法。首先从原噪声图像和其中值滤波图像得到细节图像。通过使用一种新的噪声检测方法得到另一幅图像，使其只保留细节图像中的噪声。通过这个图像，可以更加准确地检测出污染图像中的噪声。对噪声图像中的每个像素，相应滤波器输出为原像素灰度值和窗中像素中值的线性组合。当前像素是一个脉冲的可能性越大，滤波过程中对它改变的就越多。与其它的中值类滤波方法相比，该方法不仅可以有效地去除噪声，而且更好地保留了图像细节。

关键词: 脉冲噪声, 中值滤波, 噪声率