摘要
针对目前全变分模型不能在去除噪声的同时有效保持纹理信息的问题,提出了一种新的基于纹理结构的超声图像自适应去噪模型.该模型首先使用纹理信息来描述超声图像的斑点特性.根据纹理特性来定义均匀性值,从而把超声图像从灰度域映射到均匀性域.然后根据二维均匀性直方图来确定阈值从而将像素点分入均匀点集或非均匀点集.最后根据像素点所隶属的集合自适应的选择不同范数的全变分去噪方法,通过大量实验验证了所提模型的有效性.
To resolve the inability of the existing total variation （TV） model to effectively preserve textural information when denoising, a novel adaptive ultrasound image denoising model based on texture structure was proposed to improve it. First, the speckle characteristics of ultrasound images are described by textural information. A homogeneity value can be defined in terms of texture in medical ultrasound images, and the gray-scale domains of these ultrasound images are mapped to the homogeneity domain. Then a threshold is obtained by a two-dimensional histogram of homogeneity, and on this basis, different pixels are partitioned into the homogeneity set or the non-homogeneity set. Models with different norms are then adaptively chosen according to different sets. Extensive denoising of ultrasound images showed that the proposed model is superior to existing models.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2009年第11期1268-1272,共5页
Journal of Harbin Engineering University
基金
中国博士后基金资助项目(20060400809)
黑龙江省青年科技基金资助项目(QC06C022)
哈尔滨工程大学基础研究基金资助项目(HEUFT05068
HEUFT07022
HEUFT05021)
关键词
超声图像
图像去噪
纹理结构
全变分
ultrasound image
image denoising
texture structure
total variation
