复杂纹理图像的噪声评估技术研究  

Research on noise assessment algorithm of complex texture images

作　　者：王帅[1] 于劲松[1] 姜杨[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《激光与红外》2014年第5期567-571,共5页Laser & Infrared

基　　金：航空科学基金项目(No.20100751010)资助

摘　　要：现有方法对复杂纹理图像的噪声评估误差较大,本文提出一种准确估计复杂纹理图像噪声大小的方法。其主要原理是先经高通滤波器去除纹理;再对滤波后的图像进行局部标准差计算并将得到的一系列的标准差划归到整数区间进行个数统计绘制直方图;进而根据理论上的卡方分布构造修正曲线对得到的直方图曲线进行修正;最后进行迭代计算得到噪声的估计值。仿真实验结果表明:与已有的噪声评估算法相比,本方法对噪声的评估不仅精度高,而且稳定性好。Since common methods for calculating the noise of complex texture images have large errors,a method to ac-curately estimate the noise in complex texture images is proposed.Firstly,high-pass filter is used to filter out the con-tour information.Secondly local standard deviation is calculated based on filtered image,then a series of standard devi-ations are classified to the nearest whole number intervals in order to draw the histogram.Thirdly,the histogram is cor-rected according to the chi-square distribution.Finally,the noise value is calculated by iterative calculation.The simu-lation results show that compared with the existing algorithm,the noise calculated by this method has higher accuracy and stability.

关 键 词：复杂纹理 噪声方差 评估 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]