Abstract: A new adaptive median filter algorithm was proposed.By a 3×3 window sliding in the original image the block uniformity of the window was calculated and was compared with the block uniformity of the entire image，therefore，the window center pixel was adaptively determined whether it was a noise；The number of noises in the 3×3 window was counted so that the filter window size was adaptively adjusted；The weighted value of each pixel in the filter window was adaptively calculated and the improved median filter method was used to eliminate noise pixels.This algorithm can effectively remove noises and also retain image details very well.According to the further analysis of experimental results，it is shown that the proposed algorithm is superior to the traditional mean filter algorithm and median filter algorithm，under the same intensity of pepper and salt noises，with 9.4 to 12.7 improvement of PSNR．The evaluation results coincide with the visual effect very wel1.It also provides a new way for image filtering.

Key words: adaptive, median filter, block uniformity, PSNR

摘要： 提出了一种新的自适应中值滤波算法，首先使用3×3窗口在图像上滑动，计算该窗口中心像素的块均匀度，并与整幅图像的块均匀度比较，自适应地确定窗口中心像素是否为噪声点；然后统计3×3窗口中噪声点的个数，自适应地调整滤波窗口大小，最后自适应地计算权值，并采用改进的加权中值滤波方法对噪声点进行逐点滤波。该方法既能有效地去除图像噪声点，又能较好地保持图像细节部分。通过对实验结果进一步分析，该方法比均值滤波和中值滤波的性能更加优化，在椒盐噪声大小相同的情况下，PSNR值提高了9.4~12.7。评价结果与目视效果吻合良好，为图像去除噪声提供了一个新的途径。

关键词: 自适应, 中值滤波, 块均匀度, PSNR