基于相位的C-V模型乳腺超声图像分割方法  

Breast ultrasound image segmentation method using C-V model based on phase

作　　者：粟华[1] 杨冠羽[1] 胡轶宁[1] 舒华忠[1] 

机构地区：[1]东南大学影像科学与技术实验室,南京210096

出　　处：《东南大学学报（自然科学版）》2013年第3期494-497,共4页Journal of Southeast University：Natural Science Edition

基　　金：国家重点基础研究发展计划(973计划)资助项目(2011CB707904);国家自然科学基金资助项目(31100713;60911130370;61073138;61271312;81101104);教育部博士点基金资助项目(20110092110023);江苏省自然科学基金资助项目(BK2012743);江苏省"六大人才高峰"资助项目

摘　　要：为了提高乳腺超声图像分割的准确率,提出了一种基于相位特征的C-V模型超声图像分割方法.首先,采用LOG-Gabor滤波器对超声图像进行6个不同方向的滤波,提取最大能量所对应的相位信息,得到超声图像的相位特征.然后,采用SRAD方法对超声图像降噪,并将降噪后的图像与相位特征点乘,增强图像目标与背景的对比度.最后,运用C-V模型的分割算法识别图像中的目标区域,并采用腐蚀方法使目标区域边缘完整、平滑.实验结果表明,与基于灰度的C-V模型、GAC模型以及基于相位特征的人工神经网络方法相比,利用该方法分割乳腺超声图像,分割的精确度明显提高,达到92.40%.In order to improve the accuracy of breast ultrasound image segmentation,an ultrasound image segmentation method using the C-V（Chan-Vese） model based on phase is proposed.First,the ultrasound image is filtered by LOG-Gabor filters in six different orientations,and the phase feature of the image is obtained by extracting the phase information in the orientation with the maximum energy.Then,the SRAD（speckle reducing anisotropic diffusion） method is used to reduce the noise of the ultrasound image,and the processed image is multiplied by the phase features to enhance the contrast of the target and background.Finally,the target of the ultrasound image is identified by the segmentation algorithm using the C-V model,and corrosion is applied to make the edge smooth and complete.The experimental results show that compared with the C-V model and GAC（geodesic active contour） model based on image gray and the ANN（artificial neural networks） method based on phase feature,the proposed method can obviously improve the accuracy of breast ultrasound image segmentation,which is 92.40%.

关 键 词：相位特征 超声图像分割 C-V模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]