基于Laplace谱结构特征和局部特征的图像分类  

Image classification based on Laplace spectra structure feature and local feature

作　　者：孙山[1] 梁栋[1] 鲍文霞[1] 张成[1] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039

出　　处：《安徽大学学报（自然科学版）》2015年第6期60-66,共7页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61172127);安徽省自然科学基金资助项目(1208085QF104);安徽省高校优秀青年人才基金资助项目(2012SQRL017ZD)

摘　　要：鉴于常规词袋模型中图像局部特征对图像信息表达不全面的特点,提出一种基于图像Laplace谱结构特征与局部特征相结合的图像分类方法.在提取基于图像均匀划分的Laplace谱结构特征后,对图像进行尺度不变特征变换(scale-invariant feature transform,简称SIFT)的抽取及描述;构造基于图像特征的视觉词典;根据视觉词典对图像特征进行量化,得到图像的全局特征直方图表示;构造支持向量机(support vector machine,简称SVM)分类器并进行图像分类.实验验证了该方法对图像进行分类的有效性与可行性.In the conventional bag of words model, in view of the characteristic that the images＇ local features can not express images＇ information completely, a method of image classification based on the coalescent of Laplace spectra structure features and local features had been presented. Before the images＇ scale invariant feature transform （SIFT） features were extracted, we had extracted Laplace spectrum structure features based on the evenly divided images; after that, the visual dictionaries were constructed based on image features which were extracted in last step; then the image features were quantified according to the constructed visual dictionaries and the global features ＇ histogram representations of the images would be obtained. At last, we constructed a support vector machine （SVM） classifier to complete the image classification. The feasibility and effectiveness of this method to classify images had been validated by several experiments.

关 键 词：谱结构特征 局部特征 视觉词典 词袋模型 图像分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]