为有效抑制图像噪声，提高电气设备红外诊断的准确性，采用基于小波系数尺度间相关性和双变量收缩函数的方法对电缆瓷套终端红外图像进行去噪． 将图像进行小波
分解，计算小波系数尺度间的相关系数，使用模糊c-均值聚类法对相关系数聚类，即将小波系数分为有效系数和无效系数两类． 对无效小波系数直接进行置零处理，对有效小波系数使用双变量收缩函数进行处理，得到真实图像小波系数的估计值． 最后，对处理得到的真实图像小波系数的估计值进行重构，便得到去噪后图像． 含噪图像的去噪结果表明，运用文中方法能有效地去除红外图像中的噪声，且与使用传统软阈值方法去噪得到的图像对比，文中方法去噪后的图像信噪比更高，最小均方误差更小．

In order to effectively suppress the noise of images and improve the accuracy of infrared diagnosis of electrical equipment,a denoising method based on both the wavelet coefficients’inter-scale correlation and the bivariate shrinkage function is proposed to denoise the infrared images of porcelain sleeve cable terminals.In this method,first,the wavelet transform coefficients are separated into two sorts by means of fuzzy c-means clustering according to the calculated inter-scale correlation coefficients of wavelet coefficients,namely,the efficient coefficients and the invalid coefficients.Then,the invalid wavelet coefficients are directly set to zero,while the efficient wavelet coefficients are processed with the bivariate shrinkage function.Thus,the estimated values of image's wavelet coefficients are obtained.Finally,the estimated wavelet coefficients are used to reconstruct a denoised image.The denoising results of infrared images with noise show that,as compared with the traditional soft thresholding method,the proposed method is more effective because it improves both the signal-to-noise ratio and the mean square error.