【作者】 项方云；

【导师】 胡同森； 田贤忠；

【作者基本信息】 浙江工业大学 ， 计算机应用技术， 2007， 硕士

【摘要】 医学图像处理是计算机领域的一个分支,是数字图像处理技术在生物医学工程中的重要应用。对医学图像进行目标轮廓提取,是临床诊断和科学研究的基础。活动轮廓线模型即Snake模型所特有的高度动态特性,使之非常适合于医学图像的分割和轮廓提取处理。本论文具体研究使用Snake模型解决医学图像目标轮廓提取所面临的关键技术问题。把Snake模型应用于临床系统中,首先要解决的是初始轮廓线和参数难以确定的问题。本文抛弃使用工作人员手工勾勒初始轮廓的原始方法,提出结合基于标识分水岭算法的初始轮廓线获得方法,该方法能快速有效地得到趋近于实际边缘的初始轮廓线,为后面的模型迭代收敛打下基础。在设定参数上,根据活动轮廓模型中不同参数在不同时期的影响作用,预测估计非边缘点应调整到的目标位置和能量,求得下一平衡状态下参数数值从而达到自适应地调整参数的效果,使模型轮廓能更好地收敛于目标边缘。把目标的先验知识融入到活动轮廓模型中,是活动轮廓模型的发展趋势。本文结合医学图像本身的特点,使用曲线拟合的方法构造椭圆力场,使活动轮廓模型在形变过程中得以保持椭圆的形状,在分割椭圆类目标轮廓时能使模型在弱边缘处也得到很好的响应。本论文通实验的手段证明了上述方法的有效性,并能快速简单地应用于临床系统之中。更多还原

【Abstract】 As one research branch in the field of computer vision, medical image processing is an important application of computer graphics and digital image processing in the biomedical engineering. Extracting the object’s contours from medical images is the foundation of diagnostic and sceince research. Active contour models (ACM) are very suitable to medical image segmentation and contour extracting due to its highly dynamic quality. The primary task of the thesis is to study the key points of applying ACM to solve problems of precessing medical images.The first problem to be solved is how to confirm the initial contour and the parameters of the models. The thesis uses the method of initial contour selection of snake model based on watershed algorithm instead of the method of depicting the contours by manual. Result shows that the algorithm can quickly get the availabe contour that works well in the further process. For the parameters, we get the parameters by estimating the position and energy of non-edge points in the next balance state, so the parameters can be changed adaptively in order to get a better segmentation. Combining the prior knowledge of the medical images to the snake models is one of developing trends. This thesis analyse the characteristic of the medical images and improve the snake models with an extra elliptic force which is computed by elliptic curve fitting, which can maintain contours with the elliptic shapes during the converging process.The experiments followed show that the methods above work well in the medical images and can be easily applied in clinic systems.更多还原