垃圾图像判别中的特征提取与选择研究  

Research of feature extraction and selection for spam image identification

作　　者：程红蓉[1] 秦志光[1] 万明成[1] 曾志华[1] 

机构地区：[1]电子科技大学计算机科学与工程学院,成都610054

出　　处：《计算机应用研究》2009年第6期2001-2003,共3页Application Research of Computers

基　　金：国家高技术研究发展计划资助项目(2006AA01Z411)

摘　　要：对垃圾图像判别问题中的特征提取和特征选择研究现状进行了总结。从特征的可区分性、鲁棒性和提取效率三个方面比较了垃圾图像判别中的主要特征,分析了特征的优缺点。结合分类学习算法、仿真实验结果,对已有的主要特征选择和分析方法进行比对,为进一步研究特征提取、特征选择方法,提高垃圾图像分类器的性能和效率提供有价值的参考。This paper was a review of the methods of feature extraction and feature selection for spam image identification. It compared the characteristics of the extracted features based on discriminability, robustness, and efficiency. Explained the strengths and weaknesses of different features. Combined with the related classification learning algorithms and experimental resuits, analyzed the feature selection methods. This survey is a valuable reference for further research on feature extraction and selection to improve the performance of spam image detection.

关 键 词：垃圾图像 特征提取 特征选择 分类器 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]