网络环境下多尺度活动轮廓模型动态分割技术研究

项目名称： 网络环境下多尺度活动轮廓模型动态分割技术研究

项目编号： No.61202364

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 沈晓红

作者单位： 山东财经大学

项目金额： 23万元

中文摘要： 网络环境下动态分割是远程医疗系统的基础和关键环节。而精度和速度是目前该课题最具挑战的问题。本项目以边缘能量获取和形变运动估计为切入点，基于多尺度几何变换和活动轮廓模型，对含噪、弱边界医学图像动态分割中的关键问题进行系统研究。包括：1、基于多尺度几何变换，研究图像边缘能量的提取，实现多尺度边缘能量的准确获取，提高抗噪能力；2、融合多尺度边缘能量，构建多尺度活动轮廓模型，改善活动轮廓模型的边界泄漏问题，实现关键帧图像的准确分割；3、针对人体器官，基于关键帧分割结果和形变运动估计，研究图像序列的动态分割技术，实现低信噪比、弱边界、伪影图像序列的准确、快速分割；4、根据各方向子带间去相关性，研究动态分割中各模块关系，实现网络环境下动态分割的并行与分布式计算。本项目研究成果，将为网络环境下医学图像动态分割研究提供新的理论和算法，为相关应用领域中分割问题的解决提供一系列实用、快速和鲁棒的新技术。

中文关键词： 多尺度几何变换；方向窗；边缘；自相似性；变形跟踪

英文摘要： Dynamic segmentation for medical image series in network environment is one of the basic and key issues in telemedical system. Accuracy and speed are both the most challenging problems. Medical images usually have noise, blurred edges and motion artifacts. To obtain the accurate boundaries of objects rapidly, the research firstly focuses on the acquisition of edge energies and the deformation estimation of human organ and tissue, then studies the key problems in dynamic segmentation based on the multiscale geometric transform and the active contour model for medical image series systematically. The research work includes: 1. The acquisition of edge energies is studied based on multiscale geometric transform to extract edge energies from multiscale subbands of noisy images efficiently; 2. Combined by multiscale edge energies, the multiscale active contour model is constructed to avoid the boundary leakage and obtain the accurate boundaries in key frames; 3. Based on the deformation estimation and the segmentation results of key frames, the precise and rapid dynamic segmentation for image series is explored to locate objects precisely and rapidly; 4. According to the decorrelation performance among directional subbands, the relationship of modules in dynamic segmentation is discussed to implement the parallel and

英文关键词： multiscale geometric transform；directional window；edge；self-similarity；deformation tracking