谱聚类 java,基于Nystrom谱聚类图像分割算法研究.docx

PAGE \* MERGEFORMATIV

摘 要

谱聚类作为一类新兴的、有效的聚类方法得到广泛的关注，并已成功的应用于图像分割。本文研究基于谱聚类的图像分割技术，并成功的应用到图像分割中。本文主要工作如下:

谱聚类算法是利用图像的相似性矩阵进行图像分割，如何构造一个对图像信息表达更充分的相似性矩阵，对算法性能有很大影响。本文应用Nystrom谱聚类算法进行图像分割技术的研究，而且还具有良好的方向分析能力，它能反映出图像在不同分辨率上沿多个方向的变化，能更好地描述图像的几何结构。

该方法为的是在聚类的时候取得较好且稳定的性能，然而KHM仅在对低维数样本聚类时较KM算法要好。为了缓解谱聚类对初始化敏感的问题，本文采用了基于优化策略的K均值算法，实现了基于Nystrom谱聚类的图像分割方法。通过在Nystrom谱聚类算法仿真实验表明:本文的方法无论在细纹理、粗纹理、非均匀纹理和混合纹理上，其视觉效果和评价指标上都要优于的方法。

关键词:图像分割;谱聚类；Nystrom谱聚类算法；K-Means聚类算法

Abstract

Spectral clustering as a class of new and effective clustering method has beenwidely concerned and successfully applied in image segmentation. The imagesegmentation algorithm, based on initialization-independent multi-parameter kernelspectral clustering which was raised by Ma Xiuli, has been researched, improved andsuccessfully applied in texture image and SAR image segmentation in this dissertation.The main achievement of this dissertation can be summarized as follows:

Spectral clustering uses the image similarity matrix in image segmentation.How to construct a similarity matrix which expresses more information of image has agreat influence on algorithm performance. It can reflect that image changes on different resolutions along several directions, anddescribe the geometry of image better.

K-means based on optimization strategy has been adopt to realize the imagesegmentation based on complex wavelet feature and spectral clustering. A large numberof simulation experiments in Brodatz library show that: In terms of fine texture, roughtexture, non-uniform texture and blending textures, our method is better than themethod according to the visual effects and evaluation indicators.

Keyword: Image Segmentation；Spectral Clustering Dual-Tree Complex；K means Spectral Clustering

目 录

TOC \o "1-3" \h \z \u HYPERLINK \l "_Toc422490453" 1 绪 论 PAGEREF _Toc422490453 \h 1

HYPERLINK \l "_Toc422490454" 1.1数字图象处理技术 PAGEREF _Toc422490454 \h 1

HYPERLINK \l "_Toc422490455" 1.1.1 简介 PAGEREF _Toc422490455 \h 1

HYPERLINK \l