一种快速并行中值滤波算法的实现

A Fast Parallel Median Filtering Algorithm Implemented

介针对传统中值滤波算法计算量大、耗时较长的缺点提出了一种快速并行中值滤波算法。对于3×3滑动窗口,窗口的9个数据是并行传给计算比较模块的,第二级的计算也是并行进行的。新算法有效减少了重复比较操作的执行,同时也大幅减少了比较次数。对于3×3滑动窗口,新算法的比较次数为13次,相对于传统中值滤波的30次,比较次数少了2倍多。比较次数的减少,意味着算法复杂度的降低和图像滤波处理速度的提高。并且通过仿真实验可知,对于椒盐噪声的滤除,算法能够达到与中值滤波同样的视觉处理效果。

Due to the large computing capacity and time consuming disadvantage of traditional median filter algorithm, a fast parallel median filter algorithm is proposed. For 3×3 sliding window, the nine data of this window is passed to the calculating comparison module parallelly, and the second level of calculation is also carried out in parallel. The new method reduces the repeat comparison operation effectively, at the same time, decreasing the comparison times substantially. For 3×3 sliding window, the comparison times is 13 which has decreased by more than two times, compared to the traditional median filter which needs 30 times. The reduction of comparison times indicates the lower of complexity of the algorithm and the improvement of speed of image filter processing. The experiment shows that due to the filtering of salt-and-pepper noise, the proposed method can achieve the same visual effect of median filter.