【作者】 成金勇；

【导师】 范延滨；

【作者基本信息】 青岛大学 ， 计算机软件与理论， 2005， 硕士

【摘要】 图像边缘检测是许多图像处理任务的重要步骤之一。在虚拟手术系统中,对人体切片数字图像的边缘检测是关键的一个步骤,是下一步进行三维建模的基础。但是图像中边缘与噪声都分布在高频区域,而且物体的几何边缘、表面纹理等均表现为边缘,不同的图像处理任务需要提取不同的边缘信息,这些因素使边缘检测非常困难。 本文在深入分析现有边缘检测方法的基础上,重点研究了基于小波变换的边缘检测方法、GVF(梯度矢量流)Snake动态轮廓模型。对这两种方法中存在的几个不足进行了改进,并结合小波分析与GVF Snake模型两种方法的优点,提出了一种新的边缘检测方法。 论文主要内容包括: 1.分析并程序实现了基于微分算子的传统边缘检测方法,基于小波变换的多尺度边缘检测方法。对基于小波变换的边缘检测方法中出现的双边缘问题、边缘线间断等问题提出了改进算法。 2.深入研究了传统Snake动态轮廓模型、GVF Snake动态轮廓模型的理论、方法。指出了GVF Snake动态轮廓模型的不足之处,即GVF Snake动态轮廓模型不能成功逼近由两次内凹形成的“瓶型”轮廓。本文通过修改GVF力场,对GVF Snake模型进行了改进。新的算法保留了GVF Snake模型的所有优点,并且能够自动逼近图像中的“瓶型”轮廓。 3.对GGVF Snake动态轮廓模型进行了分析。指出了它的一个不足之处,即不能够逼近偶数个像素间隔的深凹型轮廓。本文通过对图像力场的处理,使得动态轮廓模型对奇数与偶数个像素间隔的深凹型轮廓都可以成功逼近。 4.结合小波分析理论和GVF Snake动态轮廓模型方法,提出了一种图像边缘检测的新方法:WVF(小波变换矢量流)Snake模型。该方法克服了小波分析方法得到的边缘不连续的缺点,同时比GVF Snake模型具有更好的抗噪性,提高了动态轮廓模型方法检测复杂图像边缘线的能力。更多还原

【Abstract】 Edge detection is an indispensable step before image analysis and recognition, and is an important technology in the image preprocessing procedure. To the surgery simulation system, the edge-detection of the human body section image is a crucial process too.This thesis discusses the edge-detection method based on the wavelets analysis, and the GVF (Gradient Vector Flow) Snake deformable model. The thesis modifies some shortages of the two methods. And based on the advantages of the two edge detection methods, a new edge-detection deformable model is put forward: WVF (Wavelets Vector Flow) Snake model.The main achievements of this thesis are as following:1. The thesis analyzes and realizes the traditional edge detection method that based on the differential operator and based on the wavelets transform. The method of wavelets transform has some shortcoming such as the edge is disconnected and is not single. To this, the thesis put forward modified algorithm.2. GVF Snake model is an efficient method for contour detection of objects on an image. In the thesis, a detailed analysis of GVF Snake model is given with emphasis on its limitations and shortcomings, and then the model is improved and a new method is proposed. The new GVF Snake model can solve the problems that the GVF Snake model can not detect the edges like a bottle.3. Based on the analysis of the GGVF (Generalized Gradient Vector Flow) Snake model, one of its shortcomings is pointed out. To the long, thin boundary indentation of the images, if the number of the indentation’s pixel is even number, the GGVF Snake model can not approach the edge. Through the disposal to the force field, the modified method is put forward.4. A new images edge-detection method is put forward: WVF (Wavelets Vector Flow) Snake model. The method solves the problem that the edge based on the wavelets method isn’t consecutive. It develops the GVF Snake model’s ability that withstands the images’ noises. The experience indicates that the new method can develop snake model’s ability to manage the complicated images.更多还原