学科分类号 110
黑龙江大学
本科学生毕业论文
题 目 超像素分割技术研究 The Research of Superpixel segmentation 姓 名 李绵绵 学 号 2011027048 院 （系） 理学院 专业、年级 数学与应用数学11-1班 指导教师 张太发 2015年6月12日
摘 要
超像素分割是计算机视觉中的热点问题，超像素的使用能有效地减少图像局部信息的冗余，大大降低了图像处理的复杂度和运算量，而且还保留了对图像进行进一步处理的有效信息，超像素技术越来越广泛地应用于许多领域。
本文对常用的几种超像素分割算法进行深入比较分析验证，采用过分割方法得到的超像素作为图像目标识别的基本单元,对现有的超像素分割方法进行研究学习，了解各个方法所要阐述算法的基本思想，以及分割步骤，找出不同算法模型分割效果的优缺点，取长补短，综述不同算法的适用情况。依据超像素分割算法模型的实现步骤对其进行预处理，得到预处理的图片。选用经过比较分析后的方法分割图像，根据实验结果，结合实际优化模型分割。
关键词 超像素 图像分割 区域合并
I
Abstract
The superpixel segmentation is the hot problem in computer vision. The usage of superpixel can reduce the redundancy of the local information in image efficiently and decrease the complexity of image processing largely. It keeps the information, which is usage in the next step image processing. More and more fields apply this technology.
The paper analysis some classical and commonly kind of algorithm in superpixel segmentation. We research some existing superpixel segmentation algorithm, whose basis pixel is based on the superpixel coming from over-superpixel segmentation. The basis idea and the segmental process are introduced in the paper and look for advantages and disadvantages among kinds of methods. We analysis the appropriate conditions in different methods. The preprocessing image is gotten by the model of superpixel segmentation. The simulation is shown by the method which is compared in the paper. Then the result is combined with the real optimal model.
Keywords superpixel; image segmentation; regions merging
II
目 录
摘 要 ........................................................................................................................................... II Abstract ...........................................................................................................................................I
第1章 绪 论 ........................................................................................................................... 1
1.1 论文背景和意义 ............................................................................................................. 1
1.1.1 选题的背景 .......................................................................................................... 1 1.1.2 研究此课题的意义 .............................................................................................. 1 1.2 国内外研究状况 ............................................................................................................ 1 1.3 本文的主要工作 ............................................................................................................ 4 第2章 基本理论 ................................................................................................................... 5
2.1 算法的模型 .................................................................................................................... 5
2.1.1 基于图论的分割方法 .......................................................................................... 5 2.1.2 基于梯度的分割方法...........................................................................................8
2.2 几种常见的算法 ............................................................................................................ 9 2.2.1 Graph-based 方法 .............................................................................................. 9 2.2.2 Superpixel lattice 方法.................................................................................10 12.2.3
基
于
熵
率
的
方
法..................................................................................................12
2.2.4 Turbopixels 方法.............................................................................................13 2.2.5 SLIC 方法...........................................................................................................14 2.2.6 Quick Shift .......................................................................................................15 第3章 实例分析 ..................................................................................................................... 16
3.1 数据的收集与处理 ...................................................................................................... 16
3.1.1 数据集获取 ..................................................................................................... 16 3.2 图像处理效果的对比分析 ........................................................................................... 17 3.2.1 分割结果比较....................................................................................................17
3.3 提出合适的评判准则 ................................................................................................... 19
3.3.1 分割存在的问题 ............................................................................................ 20 3.3.2 超像素合并策略 ............................................................................................ 20
第4章 总结与展望 .................................................................................................................... 25
结 论 .......................................................................................................................................... 26 致 谢 .......................................................................................................................................... 27 参考文献 ...................................................................................................................................... 28
Contents
Abstra .................................................................................................. 错误！未定义书签。 Abstract ............................................................................................... 错误！未定义书签。
Chapter 1 Introduction ........................................................................................................ 1
1.1 The concept of visualization in scientific computing ........................................... 1 1.1.1 Background .................................................................................................... 1 1.1.2 significance ..................................................................................................... 1 1.2 research status ........................................................................................................ 1 1.3 main contribution ................................................................................................. 4 Chapter 2 basic theory ...................................................................................................... 5
2.1 Algorithm .................................................................................................................. 5
2.1.1 the superpixel segmentation based on graph theory .................................... 5 2.1.2 superpixel segmentation based on gradient................................................8
2.2 some classical Algorithm ........................................................................................ 9 2.2.1 Graph-based method .................................................................................... 9 2.2.2 Superpixel lattice method .............................................................................10 12.2.3the
method
based
on
entropy.........................................................................12
2.2.4 Turbopixels method .....................................................................................13 2.2.5 SLIC method .................................................................................................14 2.2.6 Quick Shift ....................................................................................................15 Chapter 3 Chapter experiments .................................................................................. 16
3.1 the collecting and preprocessing ............................................................................. 16
3.1.1 data collectin ................................................................................................ 16 3.2 contrastive analysis................................................................................................... 17 3.2.1 comparison......................................................................................................17
3.3 criteria standard ....................................................................................................... 19
3.3.1 Disadvantages ............................................................................................. 19
3.3.2 merge strategy..............................................................................................20
Chapter 4 Implementation of algorithm ........................................................................... 25
Conclusions ......................................................................................................................... 26 Acknowledgements ............................................................................................................. 27 References ........................................................................................................................... 28