基于CUDA的椒盐噪声自适应中值滤波算法  

Algorithm of Adaptive Median Filter for Salt and Pepper Noise Based on CUDA

作　　者：段群[1] 吴粉侠[1] 刘小豫[1] 

机构地区：[1]咸阳师范学院信息工程学院,陕西咸阳712000

出　　处：《航空计算技术》2013年第2期60-63,共4页Aeronautical Computing Technique

基　　金：陕西省教育厅项目资助(09JK811);咸阳师范学院基金项目资助(12XSYK072;11XSYK210;11XSYK329;08XSYK339)

摘　　要：针对传统中值滤波算法的优缺点,结合椒盐噪声的特征,提出了一种有效的自适应中值滤波算法。首先依据椒盐噪声的特征,将图像像素分为噪声和信号,然后根据窗口内噪声点的个数自适应地确定滤波窗口的尺寸,仅对噪声像素进行中值滤波。但随着噪声密度和图像规模的增大,在CPU上执行的时间显著增加。分析并利用图形处理器(GPU)的并行处理特征,并在CUDA平台中实现了算法。实验结果表明,所提出的算法能够有效地去除椒盐噪声、保留边缘和细节,并且能显著缩短计算时间,随着噪声密度和图像规模的增大,最大加速比达到6 000倍。According to excellence and shortcoming of conventional median filter algorithm,combined with the salt and pepper noise characteristics,an effective adaptive median filter algorithm is proposed.The method works in this way：Firstly,According to the salt and pepper noise characteristics,the pixels of the image are divided into noise and signal,then the size of filtering window is adaptively adjusted according to the number of noise points,Finally,an adaptive median image smoothing algorithm is used to filter noise pixels.But with the noise density and image size increases,execution on the CPU time increases significantly.This article analyses and uses parallel processing of the graphics processor（GPU）,and achieves the algorithm in the CUDA platform.The experimental results show that the proposed algorithm can not only effectively remove salt and pepper noise,preserve the edges and details,but also can reduce the time consumption significantly.As the image size increases,it can get a maximum speed up ratio of 6 000 times.

关 键 词：自适应中值滤波 椒盐噪声 计算统一设备架构 图形处理器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]