K均值算法图像分割

毕业论文(设计)

题 目 基于特征的图像分割技术

学生姓名 万亚堃

学 号 20111334044

学 院 电子与信息工程学院

专 业 通信工程

指导教师 胡昭华老师

二Ｏ一五年四月五日

1

目 录

1.绪论 ........................................................................................................................................ 4 1.1

课题研究意义 ................................................................................................................ 4

1.2图像分割技术发展概况 ..................................................................................................... 5 1.3 图像分割方法的现状 ........................................................................................................ 7 1.4论文内容 ............................................................................................................................. 9 2.基于综合特征的图像分割 .................................................................................................... 9 2.1概述 ..................................................................................................................................... 9 2.2颜色空间选取 ................................................................................................................... 10 2.3图像特征提取 ................................................................................................................... 10 2.3.1颜色特征提取 ............................................................................................................ 10 2.3.2纹理特征提取 ............................................................................................................ 10 2.4综合特征分割 ................................................................................................................... 11 3.K均值算法 ........................................................................................................................... 12 3.1原始K均值算法 ............................................................................................................... 12 3.2K均值聚类分割算法 ......................................................................................................... 13 3.2.1 聚类 ......................................................................................................................... 13 3.2.2 K-均值聚类算法的工作原理： .............................................................................. 13 3.2.3 K-means聚类算法的一般步骤： .......................................................................... 13 3.2.4 K-均值聚类法的缺点： .......................................................................................... 14 3.3．基于灰度空间的彩色图像像素聚类 ............................................................................ 14 3.4改进的k-均值聚类图像分割算法 ................................................................................... 15 3.5分割结果及分析 ............................................................................................................... 20 4.本文结论 .............................................................................................................................. 20 4.1存在的问题以及对未来的展望 ....................................................................................... 20 参考文献 ................................................................................................................................. 21 致 谢 ................................................................................................................................... 23 附一：K-均值聚类改进前的matlab源程序......................................................................... 24

2

基于特征的图像分割技术

万亚堃

南京信息工程大学电子与信息工程学院，江苏 南京 210044

摘 要：图像分割是指将一副图像分解为若干互不交叠的有意义且具有相同属性的区域。图像分割是数字图像处理中的一项关键技术，其分割的准确性直接影响后续任务的有效性，因此具有十分重要的意义。现有的分割算法在不同程度上取得了一定的成功，但是图像分割的很多问题还远远没有解决，该方面的研究仍然面临很多挑战。

本文采用改进的K均值算法进行图像分割，在颜色空间选取上也采用比较好的RBG颜色空间，对图像分别进行了颜色特征提取与纹理特征提取，最后进行了原始K均值算法与改进后的K均值算法分割图形的比较，实验结果表明本文提出的方法可以很好的从图像中分割出有意义的区域，更突出目标区域。 关键词：图像分割，颜色空间，K均值聚类。

Based on the characteristics of the image segmentation

technology

Wanyakun

3

NUIST，Nanjing 210044，China

Abstract：Image segmentation is a pair of images are decomposed into several mutually overlapping area of meaningful and with the same attribute. Image segmentation is a key technology of digital image processing，The segmentation accuracy directly affect the effectiveness of the subsequent task, So it is of vital significance. Existing segmentation algorithm in different degree, has achieved some success, but is far from solved many problems of image segmentation, the research still faces many challenges.

Image segmentation is one of the most basic and important field in image processing, is to visual image analysis and pattern recognition is the basic premise. Proposed in this paper USES the improved k-means algorithm for image segmentation, on the selection of color space is better HUV based on color space (by linear RBG color space transformation). Image feature and color feature extraction of texture feature extraction respectively, finally has carried on the original k-means algorithm and the improved k-means algorithm segmentation graphical comparison of the experimental results show that the proposed approach can be very good meaningful regions segmented from the image.

Keywords: Image segmentation, color space, k-means clustering.

1.绪论

1.1 课题研究意义

图像分割是数字图像处理中的一项关键技术，它通常用于对图像进行分析、识别、编码等处理之前的预处理环节，其分割的准确性直接影响后续任务的有效性，因此具有十分重要的意义。自上世纪70年代以来，已经出现了多种图像分割方法，而每一种图像分割方法都

4