摘要
在空间光通信中,光斑图像噪声直接影响质心精度。利用小波的方法估计出这种噪声属于广义的高斯噪声,因此,提出了一种基于Bayesian估计的小波降噪方法。该方法以“Haar”函数为小波基,对带噪图像作正交小波变换;通过对小波系数的概率密度函数建模,用Bayesian方法估计无噪原始图像的小波系数;用估计出的小波系数重构得到消噪图像。实验表明,对同样的光斑图像,中值滤波降噪法质心偏离1.8μm;全局阈值法小波图像消噪法质心偏离0.8μm;而Bayesian估计的图像小波消噪方法,峰值信噪比从-12.084 dB提高到10.048 dB,质心偏离仅0.03μm。
In space optical communication, centroid accuracy depends directly on spot image noise. Wavelet transform is used to estimate the noise, namely is generalized Gaussian noise. For this, a noise reduction method based on Bayesian estimation is proposed. With “Haar” function as wavelet basis function, an orthogonal wavelet transform is carried out for images with noises. The wavelet coefficient of original image without noise is estimated by Bayesian method through modeling PDF （Probability Density Function） of wavelet coefficient. A denoised image is reconstructed by the estimated wavelet coefficient. Experiments show that when noise reduction is carried out by median filtering, the position deviation of the centroid is 1.8μm. When noise reduction is done by global threshold method, the position deviation of the centroid is 0.8μm. When image noise reduction is carried out by Bayesian estimation method, the peak SNR is improved from -12.084dB to 10.084dB, the position deviation of the centroid is only 0.03μm.
出处
《光电工程》
EI
CAS
CSCD
北大核心
2005年第8期56-58,92,共4页
Opto-Electronic Engineering
基金
国防预研基金项目
关键词
数字图像处理
小波变换
信噪比
光斑图像
Digital image processing
Wavelet transform
Signal-noise ratio
Light spots image
