基于改进三维块匹配的X射线绝缘子图像去噪

X-ray insulator image denoising based on improved 3D lock matching

由于X射线图像噪声种类多样且随机分布,文中提出一种基于改进三维块匹配的X射线盆式绝缘子图像去噪新方法。针对三维块匹配算法协同滤波中采用的小波硬阈值法造成的伪吉布斯现象,提出一种改进的小波阈值去噪方法,克服了伪吉布斯现象,保留了更多的图像细节;针对三维块匹配算法协同滤波中采用的维纳滤波法造成的"振铃效应",提出一种基于各向异性扩散的改进卡尔曼滤波方法,避免了"振铃效应",边缘清晰,细节完整。试验结果表明,该方法极大地抑制了图像噪声,提升了X射线图像的质量,为后续故障识别提供了良好的图像基础,极大地提升了故障识别的准确率。

Since the X-ray image noises are in various types and random distribution,a new method of X-ray basin-type insulator image denoising based on improved 3D block matching is proposed in this paper.In allusion to the pseudo Gibbs phenomenon caused by the wavelet hard threshold method adopted in the collaborative filtering of the 3D block matching(BM3D)algorithm,an improved wavelet threshold denoising method is proposed,which can overcome the pseudo Gibbs phenomenon,and maintain more image details.In allusion to the ringing effect caused by the Wiener filtering method adopted in the collaborative filtering of the BM3D algorithm,an improved Kalman filtering method based on anisotropic diffusion is proposed,which can avoid the ringing effect,and maintain clear edges and complete details.The experimental results show that the proposed method can greatly suppress image noises and improve the qualities of X-ray images,which provides a good image foundation for the subsequent failure recognition and greatly improves the accuracy of failure recognition.