项目名称: 基于水平集与多尺度几何分析的三维超声数据目标提取
项目编号: No.61471225
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 无线电电子学、电信技术
项目作者: 白培瑞
作者单位: 山东科技大学
项目金额: 81万元
中文摘要: 三维医学超声成像具有显示直观、体积测量精确等优点,但是,目前临床上使用的图像分析软件对三维目标的分割与识别能力有限,制约了三维超声成像技术的实际使用。三维超声数据分析与理解的困难主要源于超声成像固有的散斑噪声和模糊边界的影响。本项目提出采用多尺度几何分析技术提取三维超声数据的边缘方向信息,克服超声图像散斑噪声、模糊边界或微弱边界的影响。然后,融合均值漂移聚类与快速变分水平集分割算法,使形变模型能够自动寻找复杂目标的封闭轮廓,实现三维目标的特征提取。本项目的研究将为三维超声数据的分析与理解提供新的理论参考,为三维超声目标的自动提取与识别提供新的思路。
中文关键词: 三维超声成像;多尺度几何分析;水平集;活动轮廓
英文摘要: The advanteages of three-dimensional ultrasound imaging lie in its visual capability and accurate volume measurement, etc. However, the limited segmentation and recognition capability of 3D objects with current clinical image analysis software restricts the application of the 3D imaging technique. The main difficuties for analyzing and understanding 3D ultrasonic data arise from inherent speckle noise and blurry edges. In this project, multi-scale geometric analysis tecnique is adopted to extract the edge direction of 3D ultrasonic data.By taking full advantage of 3D surface information, the influences of speckle noise and blurry or weak boundaries can be reduced. Then, by combining mean shift clustering with fast variational level set algorithm, close contours of 3D irregular object can be searched automatically, and the 3D object can be extracted effectively. The reasearch of this project will provide new scientific references for analyzing and understanding 3D ultrasonic data. It will also provide new ideas for extracting and recognizing 3D ultrasonic objects automatically.
英文关键词: three-dimensional ultrasound imaging;multiscale geometric analysis;level set;active contour