摘要
目前图像信噪比定义方法多种多样,对常见信噪比定义进行了分类,总结了不同定义之间的内在联系、常用解法以及使用范围。针对常见图像信噪比求解方法误差较大的问题,提出了一种基于最小二乘拟合的高斯背景噪声参数估计方法。首先根据全部灰度级估计均值和方差,然后通过迭代的方式,排除高灰度非背景像素的干扰,逐步缩小统计范围,最后选择具有最小均方误差的统计参数作为背景灰度和噪声方差的估值。仿真实验证明,与常见的"局部最小方差法"和"小于均值方差法"相比,用该方法得到的背景噪声参数计算的信噪比具有更高的精度和稳定性。
There are many kinds of definition of SNR（Signal-Noise-Ratio）in image processing. We classified the definitions of SNR and summarized the correlation, calculation and applying range of them. Since the common methods for calculating the SNR have large errors,we proposed a parameter estimating method based on least-squares fitting. Firstly, estimation was made on mean and variance according to the total gray level of histogram. Then the non-background pixels that have high gray level were excluded by iteration, and thus the statistical range was reduced step by step. At last, the statistical parameters that have least square error were chosen as the estimation of background gray level and noise variance. Emulation experiment proved that： compared with the usual＂local leaset variance method＂ and ＂below mean variance method＂, SNR calculated by the background noise parameters got from this method is more accurate and stable.
出处
《电光与控制》
北大核心
2010年第1期18-21,共4页
Electronics Optics & Control
基金
国家"八六三"计划项目(2006AA703213D)
关键词
信号处理
信噪比
噪声方差
点目标图像
最小二乘拟合
signal processing
Signal-Noise-Ratio （SNR）
noise variance
point target image
leastsquares fitting
