Bandelet变换在金属断口图像去噪中的应用  

Application of Bandelet Transform in Metal Fracture Image Denoising

作　　者：聂鹏[1] 李志农[1,2] 邬冠华[2] 

机构地区：[1]郑州大学机械工程学院,郑州450001 [2]南昌航空大学无损检测技术教育部重点实验室,南昌330063

出　　处：《噪声与振动控制》2010年第5期149-152,共4页Noise and Vibration Control

基　　金：无损检测技术教育部重点实验室开放基金(No.ZD200829003)

摘　　要：提出了一种基于Bandelet变换的金属断口图像消噪方法。在提出的方法中,首先对带噪声的金属断口图像进行四叉树分割,得到各分割区域内的最佳几何流方向,并沿Bandelet块的最佳几何方向进行曲波变换,最后利用SUREShink法计算各Bandelet块的自适应阈值,然后采用多层软阈值去除噪声,进行Bandelet逆变换重构图像。实验表明,该方法是有效的,与传统的小波子带多阈值去噪法相比,该算法不仅提高了去噪后图像的峰值信噪比(PSNR),而且具有更强的边缘保持能力。A metal fracture image denoising method based on the Bandelet transformation is proposed.In this method,the noisy on metal fracture surface is segmented in quadtree,and the best geometric flow direction in each segmentation region is obtained.Then,the band transform is completed along the best geometric direction in the Bandelet block,and the Bandelet adaptive threshold is calculated by means of SUREShink method.Finally,the noise is removed by using multi-layer soft-thresholding,and the image is reconstructed by Bandelet inverse transform.Experiment result shows that the proposed method is more effective in comparison with the traditional sub-band multi-threshold wavelet denoising method.This method raises the peak signal-to-noise ratio（PSNR） and has a stronger edge-preserving ability.

关 键 词：振动与波 BANDELET变换 断口图像 消噪 四叉树 最佳几何方向 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]