小波变换联合互信息量的水平集策略分割B超病灶  

LEVEL SET STRATEGY SEGMENTING BU FOCUS WITH WAVELET TRANSFORMATION AND MUTUAL INFORMATION AMOUNT

作　　者：杨谊[1] 喻德旷[1] 申洪[2] 

机构地区：[1]南方医科大学生物医学工程学院,广东广州510515 [2]南方医科大学基础医学院,广东广州510515

出　　处：《计算机应用与软件》2016年第1期211-215,共5页Computer Applications and Software

基　　金：广东省科技计划项目(2010B060300001)

摘　　要：目前水平集方法在B超病灶分割中的应用得到很多关注,其中区域水平集模型能够获得较好的分割结果,但仍然存在对初始条件敏感和不能准确处理灰度变化非匀质的不足。针对这两个问题,运用小波变换方法分解图像,对高频子图进行基于小波模极大值的边缘检测,获得高质量的初始曲线;然后引入互信息量的概念,定义图像局部互信息量来更准确地表现轮廓线的局部特征,以引导曲线合理形变和位移。实验结果表明,水平集策略能够准确处理病灶的模糊边界,具有一定的抗噪能力。水平集策略与其他区域水平集模型相比,在定位精度和执行效率方面都有不同程度的优化。The application of level set methods in B-ultrasound （BU） focus segmentation have gained much attention at present, and among them the regionM level set model can get better segmentation result than other types, but there are still some shortcomings in it, including being sensitive to initial condition and unable to deal with inhomogeneous intensity. Targeting at these two deficiencies, we decompose the image by wavelet transformation, and detect the edges based on wavelet modulus maxima for high frequency sub-image to acquire high quality initial curve. Then we introduce the concept of mutual information amount to define image local mutual information amount so as to more accurately express the local properties of contour, and to guide reasonable deformation and motion of curve. Experimental results show that the proposed level set strategy can precisely deal with blurred edges of focus, and has certain noise resistance capability. Compared with other regional level set models, the proposed model optimises in both locating accuracy and executing efficiency with different degrees.

关 键 词：区域水平集 小波变换 图像局部互信息量 B超病灶分割 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]