嵌入式系统实时中值滤波的高效实现

High Efficiency Implementation of Real-Time Median Filtering in Embedded System

为了满足嵌入式系统实时数字滤波的需要,针对抗脉冲干扰能力较好的中值数字滤波,提出一种存储量小、运算速度快的高效算法。该算法利用循环存储结构,在存储实时新数据的同时,自动剔除最老的旧数据,释放其所占用的存储资源,避免为新到来的数据申请存储资源。利用排序链表存储中值滤波的原始数据,避免了其他中值滤波算法在排序时必需的数据移动或交换,使中值滤波算法中的排序计算量降低到O(N)级,提高了算法效率。实测对比表明,提出的中值滤波算法的存储器使用量至多是传统中值滤波算法的37%,运算速度至少是Matlab函数库medfilt1()给定算法的2倍。

To meet the requirements of real-time digital filtering in embedded system,a high efficiency digital algorithm small storage occupancy and fast computation speed was proposed for median filtering.In the proposed algorithm,memory applying and releasing operations in data updating were reduced to the lowest limit by using cycle storage structure,which stores newly incoming real-time data to the memory position taken by the oldest data to be deleted.Link-list was borrowed to keep the order of sorted data in the cycle storage structure,which avoided data exchange in sorting procedure,which is necessary in other median filtering algorithms,and improved efficiency of the algorithm by reducing computation complexity of the proposed median filtering to be O（N）.Comparative experiment showed that the proposed median filtering algorithm at most took up 37% storage space of the traditional one and ran at least 2 times faster than Matlab medfilt1（）function does.