【作者】 康颐；

【导师】 孙淑霞；

【作者基本信息】 成都理工大学 ， 计算机应用技术， 2008， 硕士

【摘要】 数字图像的边缘检测是图像分割、目标区域识别、区域形状检测等图像分析领域十分重要的基础,是图像识别中检测图像特征的一个重要属性,图像理解和分析的第一步往往就是边缘检测,目前它已成为机器视觉研究领域最活跃的课题之一。医学CT图像与普通图像比较,本质上具有模糊性和不均匀性。首先,图像具有灰度上的含糊性,在同一种组织中信号值会出现大幅度的变化。其次,由于局部体效应,一个边界上的体素中常常同时包含两种物质。另外,由于技术上的原因带来的噪声往往模糊了物体边缘的高频信号。本文主要研究基于小波变换的医学CT图像边缘检测方法。首先概述图像边缘检测的基本原理,介绍了当前国内外对于边缘检测典型研究方向;接着介绍了小波变换及其应用于图像边缘检测的相关理论;最后重点研究了基于小波变换的多尺度图像边缘检测算法,对比了多种基于小波变换的多尺度图像边缘检测改进算法,并对它们的检测效果进行了比较。在此基础上,结合两种改进算法的优点提出了一种基于小波变换的多尺度自适应边缘检测改进算法,改进算法的目的是使医学CT图像边缘检测具有更良好的噪声抑制能力,同时又具有完备的边缘保持特性,并进行了相关实验验证,以验证该改进算法的合法性和合理性。更多还原

【Abstract】 Many fields related to image analysis are based on the edge detection of digital images, such as image segmentation, recognition of target region, shape extraction from regional areas and so on. The edge detection of digital image is an important attribute for extracting image characteristics in image recognition. It is also the first step in image understanding and image analysis. It has become one of the most active issues in the field related to machine vision research. Comparing with ordinary images, medical CT images are often fuzzy and with heterogeneity. First, they have ambiguities in gray-scale, signal values coming from one tissue will have drastic changes. Secondly, because of the partial effect, some places on the border of the tissue often contain two substances.In this paper, we set focus on the methods of detecting medical CT image edges based on wavelet transform. We firstly summarize the basic theory on edge detection of images, and introduce some new domestic and foreign research directions on this field; Then, we introduce the theory on wavelet transform and the theory of edge detection of image based on wavelet transform; At last, we mainly study multi-scale methods of detecting image edges based on wavelet transform, citing several improved methods, comparing the results of these methods. Then we propose an improved adaptive method based on two improved methods, the main objective of this improved algorithm is to enhance the quality of detecting medical CT image edges with minimal affection of fuzzy signals and all of the characteristics of the edge, and we verify the improved algorithm’s performance by doing experiments.更多还原