中值滤波在图像预处理阶段有着重要的应用，尤其是对于图像中脉冲噪声的滤除。传统的中值滤波算法都是通过对原数据按照大小排序，然后才能取出中值。但随着窗口尺寸的扩大，因数据增多而造成比较次数的激增，这成为制约大窗口中值滤波器实时性的一个瓶颈。针对这一问题，该文提出一种基于FPGA的25输入的实时中值滤波器的设计方法。无需对原数据进行大小排序，而是根据各比特平面的信息，重新生成一个中值，特别适合于FPGA实现。设计专用的位处理单元，简化了电路结构。通过三级查找表的方式，解决“多数表决器”耗时长的问题。实验结果表明，滤波器资源占用少，数据吞吐率高，延迟小，可满足任何窗口形状的滤波要求。%The median filtering has important application in image preprocessing stage,especially for filtering of impulse noise in image. The traditional median filtering algorithm is used to sort the original data according to its size to extract the medi⁃an. The enlarged window size and increasing data cause the dramatical increase of comparison times,which becomes the bottle⁃neck to restrict the real⁃time performance of big⁃window median filter. For this problem,a design method for the real⁃time medi⁃an filter with 25 inputs based on FPGA is presented. It is unnecessary to sort the original data,but a new median is generated according to the information of each bit plane,which is particularly suitable for the filtering implementation based on FPGA. A special bit⁃processing unit was designed to simplify the circuit structure. The method of three⁃level lookup table can solve the time⁃consuming issue of“majority voter”. The experimental results show that the proposed filter has less sources occupation, high data throughput and short time delay,and can satisfy the filtering requirement of any window shapes.
展开▼
