CASIA OpenIR  > 毕业生  > 博士学位论文
医学影像的分割算法及存储管理的研究与应用
其他题名Study and Application of Medical Images Segmentation, Storage and Management
葛行飞
学位类型工学博士
导师田捷
2003-05-01
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业模式识别与智能系统
关键词医学影像 图像分割 同质性直方图 主动轮廓 梯度矢量场 Mean Shift 色彩空间 区域增长 集群 Pacs Medical Lmage Image Segmentation Homogram Active Contour Gvf Mean Shift Color Space Region Growing Cluster Pacs
摘要医疗诊断设备极大地提高了人类获取自身信息的能力,推动了现代医学的发 展。这些设备获得的信息主要以图像的形式存在,因此对这些图像的处理就成 了充分发挥其效果的关键。由于其成像方式的共同特点,我们把这些来自各种 医疗诊断设备的图像统称为医学影像。 本文主要对医学影像处理中的分割算法及其存储与管理进行了研究。首先是 以算法研究为主的灰度医学影像和彩色医学图像的分割,其次是面向应用的医 学影像存储与管理方面的研究与开发。本文的主要工作包括: 图像分割算法的研究方面: 1)提出了一种用像素同质度对医学影像进行加权以改善其直方图性能,以 及快速的自动阈值确定准则来决定阈值并对医学影像进行分割的算法。针对很 多医学影像,主要是核磁共振影像的直方图峰谷不突出,因而很难用简单快速 的阈值方法分割的情况,通过计算像素的同质度,并用同质度对原始图像进行 加权后计算直方图,以改善直方图的性能。提出了一种波峰查找和筛选规则, 根据这一规则确定合适的阈值,并分割医学影像。而改善分割效果的同时,并 没有引入复杂的数学运算,完全避免了迭代和递归等计算量较大的计算,因而 保持了阈值算法高效、快速的优点。实验表明,采用该方法可以显著地改善直 方图的性能,快速准确地分割核磁共振影像。 2)对基于主动轮廓模型的方法进行了研究和改进。提出了一种新的主动轮 廓初始化方案,定义了梯度矢量场的扩散中心,并据此进行初始化。对得到的 多个轮廓,提出了基于外能的准则来辨别真伪边界。对于医学影像,首先用各 项异性滤波进行平滑预处理,以便通过Canny边缘提取得到较好的边缘图;为 避免复杂影像中各部分相互干扰,提出了基于移动窗口的梯度矢量场生成方式, 把参与叠加运算的像素限制在窗口内,从而改善了梯度矢量场,并且提高了计 算速度。对拓扑结构比较复杂的医学影像,提出了采用会字塔编码方式得到图 像的缩略图,把握图像的主要轮廓形状(Profile),在此基础上选择扩散中心, 并针对特殊拓扑结构图像采用相似性准则进行初始化的方法。实验表明,该方 法可以显著改善由医学影像生成的梯度矢量场,并较好地解决主动轮廓模型初 始化的问题,筛选后的轮廓能够分割出拓扑结构复杂的医学影像。 3)对彩色医学图像的分割进行了研究和探索。对彩色医学图像特点进行了 分析,提出了以根据感兴趣区域自适应色度域带宽的方式计算Mean ShiR
其他摘要Nowadays, the development of modem medicine profits from the medical modalities to a large extent, because these modalities can help us to see what we cannot before. As the data obtained from them are generated as images, process of these images plays an important role in utilizing them for the diagnosis. This thesis focuses on segmentation, storage and management of the medical images. The contents of this dissertation are described as follows: 1) A medical image segmentation algorithm is proposed which improved the histogram of the image by weighting the original images. An effective criterion is also presented to determine the appropriate threshold to segment the images. Homogram, or histogram based on homogeneity is employed in our algorithm. MR images are difficultly segmented via this method; as the gray levels of their pixels are too similar to distinguish. In our algorithm, the image is updated with the homogeneity weighted original and average gray levels. The homogram is calculated based on the updated image. It is much steeper than the regular one. Therefore a simple but agile peak-finding approach is able to determine objects to segment and corresponding thresholds exactly. Segmentation via thresholding is feasible now even in MR images. Moreover, our algorithm remains speedy even though the accuracy of segmentation advances. 2) We present an automatic active contours model which can extract multiple objects in an image without any manual assistance. Initialization of contours depends on some points which we called centers of divergence. These points are computed by Gradient Vector Flow field. One active contour is generated according to each center of divergence. After normal deformation, we get desired contours as well as some undesired. By an external energy criterion, these contours which locate in true boundaries are sought out. The medical image is smoothed by an anisotropic filtet before extracting the edge map from it with a canny operator. A method named mobile window is presented to compute the GVF field of the large and intricate medical images. The miniature of the original image is generated by a pyramid coding scheme to grasp the profile but details of the image. The experiments show that the performance of the GVF field from the medical images has been improved, and the selected contours have found the correct boundaries. 3) The chromatic medical images are analyzed and studied. A hierarchical method is presented to segment the color images. The mean shift based filter is employed to process the image before a region growing scheme could outline the region of interest (ROI). We update the generation of the mean shift vector by adaptively adjusting the band width of the color range kernel according to the ROI. A cylindrical metric in the HSI (Hue, Saturation, and Intensity) color space is utilized to measure the difference betw
馆藏号XWLW746
其他标识符746
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/5760
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
葛行飞. 医学影像的分割算法及存储管理的研究与应用[D]. 中国科学院自动化研究所. 中国科学院研究生院,2003.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[葛行飞]的文章
百度学术
百度学术中相似的文章
[葛行飞]的文章
必应学术
必应学术中相似的文章
[葛行飞]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。