基于改进局部保持映射的图像特征提取算法  

The Algorithm Image Feature Extraction Based on Improved Locality Preserving Projection

作　　者：刘靖[1] 

机构地区：[1]黑龙江省政法管理干部学院

出　　处：《哈尔滨师范大学自然科学学报》2010年第6期59-62,共4页Natural Science Journal of Harbin Normal University

基　　金：黑龙江省自然科学基金青年基金资助(QC2010066)

摘　　要：针对目前局部保持映射算法(LPP)因忽略类标识信息而造成的计算效率问题,提出了基于类标识信息的监督局部保持映射算法,并将其用于图像特征提取中,该算法通过用类标识信息指导最邻近图构造过程来提高算法的效率,仿真实验的结果表明,该方法可以有效提高传统局部保持映射算法的运算效率,提高图像特征提取性能.For the computation efficiency problem owing to the ignorance of class labels endured by LPP,the paper proposes class labels based supervised locality preserving projection algorithm and applies it to extract the image features.The algorithm enhances the computation efficiency through applying the class labels information to guide the procedure of constructing the nearest neighbor graph.Simulation results show that the proposed algorithm can increase the computation efficiency of the traditional locality preserving projection and enhance the performance of image feature extraction.

关 键 词：机器学习 图像处理 特征提取 局部保持映射 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]