基于信号可信度的高性能快速图像滤波  

Effective and fast image filtering based on signal credibility

作　　者：牛连强[1,2] 孙忠礼[2] 彭敏[2] 邵中[1] 

机构地区：[1]沈阳工业大学软件学院,沈阳110870 [2]沈阳工业大学信息科学与工程学院,沈阳110870

出　　处：《沈阳工业大学学报》2013年第6期680-686,共7页Journal of Shenyang University of Technology

基　　金：辽宁省科技计划资助项目(2007410003);沈阳市科技计划资助项目(F12-168-9-00)

摘　　要：为了快速、准确地检测出噪声并寻求可靠的替代信号,提出一种以对滤波窗口中的信号进行全面分类并赋予可信度等级为基础、根据可信度由高到低的顺序控制噪声检测和滤波行为的图像滤波方法.该方法从整体上加强对高可信度信号的优先检测和选用,在噪声污染较小的窗口中,当前信号的检测优先于对中值的检测;在噪声污染较大的窗口中,仅检测当前信号,此方法具有自适应调整窗口的能力,并在最近的有效信号内选择噪声点的替代值.分析和实验表明,该方法从噪声检测的准确性、噪声滤除的有效性和滤波速度3个方面都较现存算法有较大幅度提高,能够有效地对各种密度脉冲噪声污染后的图像进行滤波.In order to realize the fast and accurate detection of noises and search the reliable alternative signals, an image filtering method, where the signals in the filtering windows was comprehensively classified and assigned the different credibility levels as well as both noise detection and filtering behavior were controlled according to the credibility levels from high to low, was established. In the proposed method, the signals with high credibility are preferentially detected and selected. For those windows with slight noise pollution, the detection of current signals is prior to that of medians. For those windows with larger noise pollution, only the current signals are detected. Futhermore, the proposed method exhibits the capability of adaptively adjusting the windows, and can select the substitute values of noise points from the nearest effective signals. The analysis and experiments show that compared with the exsiting algorithms, the proposed method has significant improvement in three aspects such as the accurateness of noise detection, validity of noise removal and speed of filtering, and can effectively realize the image filtering after the pollution of impluse noises with various density.

关 键 词：脉冲噪声 中值滤波 噪声检测 自适应窗口 可信度 细节保护 去噪 结构相似性 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]