Abstract: Based on the similarity of neighborhood pixels, algorithm of grouping vote for median value is proposed in this paper. It avoids a lot of zero ballot-boxes into the statistical process and improves the rate of inspecting ballot-slips. It also proposes vote against idea. Using the correlation of windows mobile, it makes pixel number greatly reduced in every voting, and avoids the cleaning operation for ballot-boxes before vote. The voting is optimized, the voting rate is increased. Eliminating the existing high-speed median filter algorithm is not suitable for small window defect. It uses less summation and comparison operators. To make all kinds of window median filtering rate is greatly improved. It will get local parameters （the local extremes, the number of local extremes, and the local average） of the image quickly. These parameters are used as the basis for noise detection, can accurately determine noise and enhance the effect of the median filter.

Key words: image processing, median filtering, vote, grouping vote, vote against

摘要： 以邻域像素的相似性为基础，提出了分组投票取中值法，避免了大量的零值票箱参入检票过程，提高了检票速度。还提出了投反票思想，利用窗口移动的相关性，使每次参入投票的像素个数大大减少，避免了每次投票前的票箱清理操作，优化了投票过程，提高了投票速度。消除了现有高速中值滤波算法不适合小窗口的缺陷，用较少的求和运算和比较运算，使各类窗口下的中值滤波速度大幅度提高。还可以快速获得图像的局部参数（局部极值、极值的个数、局部均值），用这些参数作为噪声检测的依据，能够准确地判断噪声，提高中值滤波的效果。

关键词: 图像处理, 中值滤波, 投票, 分组投票, 投反票