基于均值移动算法的医学图像分割-生物医学工程专业论文.docx

摘要 摘要 摘要 医学图像分割是医学图像处理中一个重要环节，分割效果与计算机辅助诊断 结果息息相关。而随着科学技术的进步和新型理论的不断提出，新的图像分割算 法也如雨后春笋般涌现出来。 本文采用的均值移动（Mean shift）算法是一种非参数密度梯度估计算法，可 以应用到很多领域，如聚类，图像平滑，目标跟踪，图像分割等。目前，均值移 动算法已在医学图像分割中得到广泛应用。本文围绕均值移动算法做了以下工作： 首先，本文深入研究均值移动理论。对均值移动理论的发展做了简单地介绍， 并详细介绍了几种常用的核函数。采用均值移动滤波算法做了仿真实验，验证了 核带宽参数对均值移动滤波效果的影响。对结合均值移动和区域合并的图像分割 算法做了仿真实验。 其次，引入新的迭代步长，结合已有加速策略提出一种改进的加速算法，并 结合区域合并算法完成图像分割。 最后，改进了自适应带宽函数，将自适应均值移动滤波和蚁群聚类算法结合， 对蚁群聚类算法做了改进，减少了蚁群聚类算法的运算量，提高了算法的运算效 率。 关键词：医学图像分割 均值移动算法 自适应带宽 蚁群聚类算法 Ab Abstract Abstract Medical image segmentation plays a very important role in medical image processing, and the effect of segmentation will impact the results of computer assisted diagnosis. With the high speed development of science technology and the new algorithms successive proposed, a plenty of new image segmentation algorithms are proposed. This paper utilized a nonparametric density gradient algorithm called the mean shift algorithm which has been applied in many fields, such as clustering, image smoothing, target tracking, image segmentation etc. Currently, mean shift algorithm has gotten a great use of medical image segmentation. This paper did researches on mean shift algorithm as follows: First, this paper did an intensive study into mean shift algorithm. This paper introduced the development of mean shift algorithm and a little of common used kernel functions. In order to prove the effect of kernel function on the result of image smoothing, this paper did some mean shift image smoothing experiments. This paper did some simulation experiments with combining mean shift and region merging image algorithm. Second, this paper intruduced the mean shift iteration step, and proposed a new improved acceleration algorithm with some used acceleration method, and then, used the region merging algorithm to complete image segmentation. Last, this paper improved the adaptive bandwidth function. This paper combined the adaptive mean shift smoothing and ant clustering algorithm, and improved the ant c