TLS估计的小波自适应零值绝缘子红外热像去噪

Wavelet adaptive denoising method for faulty insulators infrared thermal image based on Total Least Squares estimation

为了弥补传统Bayes估计的小波去噪方法依赖于小波系数先验分布模型的不足,针对零值绝缘子红外图像具有低信噪比特点,提出了基于总体最小二乘(TLS)估计的小波自适应零值绝缘子红外热像去噪方法。受噪声污染的零值绝缘子红外图像经小波变换后,不处理低频小波系数,获取各尺度、各方向的高频小波系数进行总体最小二乘估计,对估计后的小波系数进行逆变换得到去噪后的图像。实验结果表明,该方法与软阈值法和Bayes估计法相比,能够有效去除噪声,保留了图像的细节信息,去噪效果良好。

In order to make up for traditional wavelet denoising based on Bayes estimation depends on prior distribution of wavelet coefficients, and on account of faulty insulators infrared image with characteristics of low SNR （Signal to Noise Ratio）, a wavelet adaptive denoising method for faulty insulators infrared image based on Total Least Squares （TLS） estimation is presented. Wavelet transform is applied to the image and keeps the low frequency wavelet coefficients unchanged. The high frequency wavelet coefficients on various scales go through TLS esti- mation before they together with the original low frequency wavelet coefficients, are used to reconstruct images. Experimental results show that the method effectively removes the noise, keeps the image details, and has a better denoising effect, compared to soft threshold and Bayes estimation.