基于GVF模型的目标识别与图像分割方法毕业论文

 2021-04-10 11:04

摘 要

随着科技的发展,计算机图像分割技术已经愈来愈趋于成熟,并在各个方面,包括医学图片识别,疾病诊断,图像可视化及计算机集成技术上得到广泛的应用。由于传统的图像分割方法,例如阙值法和边缘检测对图像的质量要求过高,而且非常容易受到外界的干扰,抗噪性比较差,所以,业界一般都采用Snake变形模型方法来解决传统图像分割方法所造成的提取边界效果不明显或者需要对提取后的图片进行去错处理的弊端。

Snake模型有着一般的图像处理方法所无法比拟的优点,其集图像数据、初始估计、目标轮廓及基于知识的约束统一于一个过程中;经适当的初始化后,它能自主地收敛于能量极小值状态;尺度空间中由初到精地极小化能量可以极大地扩展捕获区域和降低复杂性。 但是Snake变形模型也并非十全十美,它也存在着对轮廓初始线要求过于精确,必须要贴近目标区域才能得到较好的分割效果,而且对凹形图像分割效果很差的弊端。

为了解决Snake’模型的弊端,本文引进了一种新的变形模型外力来解决Snake模型这二个弊端,即梯度矢量流(GVF)模型场。相对于传统Snake模型而言,GVF模型对初始轮廓线和凹形物体的分割准确性更高。本文分析了GVF模型的优缺点和模型组成原理分析GVF力场的外力特征,得出GVF场停留在错误边界的原因。并用MATLAB代码完成了GVF的数值实现。另外,本文针对不同的图形做了大量的实验,并针对实验结果给出了自己的小结和问题的原因。最后通过对传统的GVF模型的算法分析,对GVF的发展开出一些研究的方向

关键词:图像分割技术 Snake模型 GVF模型

Abstract

With the development of technology, computer image segmentation has become increasingly mature, and in all aspects, including medical image recognition, diagnosis, image visualization and computer integration technology is widely used. As the traditional image segmentation methods, such as Que value method and edge detection on the image quality is too high, and very susceptible to outside interference and noise is rather poor, therefore, the industry generally Snake deformation model using traditional methods to solve the image segmentation caused by the extraction of the boundary effect was not obvious or need to extract the picture to go wrong after the treatment of the state.

Snake model has a general image processing methods can not match the advantages of the set of image data, the initial estimate, target profile and the knowledge-based constraints are integrated in a process; by proper initialization, it can independently converge minimum energy state; scale space refined by the early to minimize the energy can greatly expand the capture area and reduce complexity. Snake deformation models, but not perfect, it also exists on the initial line profile is too precise requirements, must be close to the target area to get better segmentation, but also on the concave poor image segmentation defects.

In order to solve the Snake 'model of the drawbacks, this paper has introduced a new external force to solve the deformation model of Snake model these two drawbacks, namely, gradient vector flow (GVF) model field. Compared with the traditional Snake model is concerned, GVF model, the initial contour and concave higher segmentation accuracy. This paper analyzes the advantages and disadvantages of GVF model and the model theory of the composition of the external features of GVF force field, draw the border GVF field to stay in the wrong reason and the principle of noise. MATLAB code with the completion of the GVF in the numerical implementation. And through the traditional GVF model deformation method - finite difference method to discuss the limitations that exist, so the topology of the GVF Snake variability made the feasibility of a simple vision.

Key words: computer image segmentation GVF model Snake model

目录

第一章 绪论 5

1.1研究背景和意义 5

1.2国内外研究概况 6

1.3本文主要工作 8

1.4论文主要安排 8

第二章 图像分割的简介和开发工具 9

2.1图像分割的定义和简介 9

2.2 MATLAB简介 10

第三章 Snake模型概述和数学建模 12

3.1 Snake基本原理 12

3.2 Snake模型的优缺点解析 14

第四章 GVF模型与程序实现 17

4.1 GVF模型的数学建模与实现 17

4.2关于GVF的优缺点概述 27

第五章 总结与展望 29

参考文献 32

致谢 34

第一章 绪论

1.1研究背景和意义

自从Kass[1]提出Snake模型以来,Snake模型在近30年间得到了广泛的应用,在医学图像分割,疾病诊断和计算机图像可视化上大放异彩。Snake模型最具有划时代的意义,毫无疑问是改变了之前一直流行的严格独立的计算机分层视觉模型。由于在使用计算机完成定位并实现一个图像的可视化的任务时,一般都采用自底向上的策略方式,比如著名的Marr视觉计算理论[2] ,就把事物分成次序严明的3层,不同的现象必须在不同的层次上进行研究和解释。本质上讲,在Snake模型以前的图像研究理论,都只能借助于原有图像本身的基础进行研究,而一旦事物图像本身有较严重的损毁或者干扰问题,比如噪声,纹理,投影或者遮挡等意外情况出现,那么原始的图像识别技术就很难发挥作用。究其原因,主要还是因为极度依赖原始资料的缘故,一旦原始资料精确度不够,那么所谓的图像分割和识别也无从谈起。

随着Snake模型的提出,这一困扰了业界多年的问题随之找到了新的发展方向。尽管除了Snake模型之外还有很多新的模型种类被提出,但是最具代表性的,还是这一经典模型。

您需要先支付 80元 才能查看全部内容!立即支付

课题毕业论文、开题报告、任务书、外文翻译、程序设计、图纸设计等资料可联系客服协助查找,优先添加企业微信。