计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (26): 174-176.DOI: 10.3778/j.issn.1002-8331.2008.26.053

• 图形、图像、模式识别 • 上一篇    下一篇

基于冗余小波变换的医学超声图像去斑噪算法

鄢 薇1,侯建华2   

  1. 1.湖北省中医学院附属医院 超声波室,武汉 430061
    2.中南民族大学 电子信息工程学院,武汉 430074
  • 收稿日期:2007-11-05 修回日期:2008-01-09 出版日期:2008-09-11 发布日期:2008-09-11
  • 通讯作者: 鄢 薇

Speckle reduction algorithm for medical ultrasound images based on redundant wavelet transform

YAN Wei1,HOU Jian-hua2   

  1. 1.Department of Ultrasonic,Hubei College of Traditional Chinese Medicine,Wuhan 430061,China
    2.School of Electronic Information Engineering,South-Central University for Nationalities,Wuhan 430074,China
  • Received:2007-11-05 Revised:2008-01-09 Online:2008-09-11 Published:2008-09-11
  • Contact: YAN Wei

摘要: 医学超声图像中固有的斑点噪声严重降低了图像的可解译程度,影响了后续的图像分析和诊断。提出了一种基于冗余小波变换的超声图像去斑算法,首先对含斑图像进行对数变换,将乘性噪声变成加性噪声;再对转换后图像做冗余小波分解;在小波系数服从广义高斯分布的前提下,计算每个小波高频子带的贝叶斯萎缩阈值,利用软阈值方法修正小波系数。实验结果表明,该算法去斑性能优于传统的空间域滤波和正交小波阈值去噪方法。

关键词: 医学超声图像, 斑点噪声, 冗余小波变换, 广义高斯分布, 贝叶斯萎缩阈值

Abstract: The inherent speckle noise in ultrasound(US) images severely degrades the image interpretation and affects the following-up image processing tasks.A speckle reduction algorithm is proposed for medical ultrasound images based on Redundant Wavelet Transform(RWT).At first,logarithmical transform is performed to original speckled US image to transform multiplicative noise into additive ones.Secondly,redundant wavelet transform is carried out to the transformed image.Under the assumption that the statistics of wavelet coefficients is Generalized Gaussian Distribution(GGD),BayesShrink threshold is calculated for each high frequency subband,and wavelet coefficients in the subband are modified via soft-thresholding rule.Experiment results show that the presented algorithm yields better despeckling performance than traditional spatial filterings and thresholding denoising algorithms based on Discrete Wavelet Transform(DWT).

Key words: medical ultrasound images, speckle noise, Redundant Wavelet Transform(RWT), Generalized Gaussian Distribution(GGD), BayesShrink thresholding