基于D-S可信度加权模型的高密度脉冲噪声识别算法  

High-Density Impulse Noise Recognition Algorithm Based on D-S Credibility Weighted Model

作　　者：马晓剑[1] 李梦娜 王金凤 MA Xiaojian;LI Mengna;WANG Jinfeng(School of Science,Northeast Forestry University,Harbin Heilongjiang 150040,China)

机构地区：[1]东北林业大学理学院,黑龙江哈尔滨150040

出　　处：《传感技术学报》2022年第6期769-777,共9页Chinese Journal of Sensors and Actuators

基　　金：中央高校基本科研业务费专项资金项目(2572018BC21);黑龙江博士后基金项目(LBH-Z18003);黑龙江省大学生创新创业训练计划项目(S202010225035)。

摘　　要：D-S证据理论在解决不确定性问题上具有明显的优势,但是用该理论识别脉冲噪声时会出现证据冲突问题,讨论了产生冲突的三种情形,并针对这三种情形提出了基于D-S可信度加权模型的高密度脉冲噪声识别算法。该算法引入冲突系数来量化证据的冲突程度,利用可信度来进一步修正证据,从而给出更为合理的噪声识别信息,最后,该算法将修正后的证据根据D-S证据理论进行融合,并使用Pignistic概率转换得到概率分布函数,完成脉冲噪声的识别。仿真实验结果表明,本文算法有效地解决了检测噪声时出现的证据高冲突问题,即使在95%的高密度噪声的情况下,仍然能进行有效的噪声识别,并在滤波后,很好地保留了图像的细节信息。D-S evidence theory has obvious advantages in solving the uncertain problem,but evidence conflict will occur when using this theory to identify impulse noise.Three situations of conflict are discussed,and a high-density impulse noise recognition algorithm based on the D-S credibility weighted model is proposed.The conflict coefficient is introduced to quantify the conflict degree of evidences,and the credibility is used to further modify the evidences to provide more reasonable information for identifying noise.Finally,the corrected evidences are fused according to D-S evidence theory,and the probability distribution function is obtained by Pignistic probability transformation,then the judgment of pulse noise is completed.The simulation results show that the proposed algorithm solves the problem of high conflicts.Even when the noise density is 95%,it can still effectively identify the noise,and retain the details of the image.

关 键 词：脉冲噪声识别 证据理论 信息冲突 可信度 

分 类 号：TP911.73[自动化与计算机技术]