基于小波系数Laplace模型的混合傅里叶-小波方法在去除电缆瓷套终端红外图像白噪声中的应用  

Hybrid Fourier-Wavelet De-noising Method Based on Laplace Model for Wavelet Coefficients for Infrared Image White Noise of Porcelain Sleeve Cable Terminal

作　　者：吴炬卓[1] 牛海清[1] 许佳[1] 

机构地区：[1]华南理工大学电力学院,广州510641

出　　处：《电瓷避雷器》2016年第3期20-24,29,共6页Insulators and Surge Arresters

基　　金：国家863计划资助项目(编号:2011AA05A120)

摘　　要：抑制图像噪声是电气设备红外诊断技术的前提。为有效抑制白噪声,本文提出一种用于电缆瓷套终端红外图像的改进混合傅里叶-小波去噪方法。该方法先在傅里叶域中采用维纳滤波器去噪,得到初步去噪后的红外图像;再针对初步去噪后的图像,在小波域中采用考虑小波系数尺度间相关性的拉普拉斯模型对小波系数建模,并在此基础上,运用最大后验概率估计估计出真实图像的小波系数,利用真实图像小波系数的估计值来重构信号,得到最后的去噪图像。数值试验表明,与传统的软阈值方法比较,运用该方法去噪后的图像具有更高的信噪比(SNR)和更小的最小均方误差(MSE)。Suppression of image noise is the premise of infrared condition diagnosis of electrical equipment. To improve the effectiveness of suppressing white noise, a modified hybrid Fourier-wavelet de-noising method used for the infrared image of porcelain bushing cable terminal is proposed. The noisy image is first processed in the Fourier domain by using the Wiener filter and obtained the infrared image of preliminary de-noising. Then in wavelet domain, the wavelet coefficients are modeled by using the Laplace model which considering wavelet coefficients inter-scale correlation and the wavelet coefficients of the real image are estimated by using MAP estimation. Finally, the processed wavelet coefficients are used to reconstruct the signal and get the final de-noising image. Numerical experiments indicate that the de-noising ability of the method in this paper is better than the traditional soft threshold for its higher SNR（signal-to-noise rate） and smaller MSE（minimizes the mean squared error）.

关 键 词：图像去噪 傅里叶变换 小波变换 拉普拉斯模型 最大后验概率估计 

分 类 号：TM24[一般工业技术—材料科学与工程] TP391.41[电气工程—电工理论与新技术]