基于判别分析的图形图像分类方法  

Classification Method for Graphic and Picture Based on Discriminant Analysis

作　　者：蔡圣燕[1] 聂旋[1] 郭淑勤 

机构地区：[1]天津科技大学包装与印刷工程学院,天津300222 [2]广东万昌印刷包装有限公司,佛山528315

出　　处：《中国印刷与包装研究》2013年第3期24-28,共5页China Printing Materials Market

基　　金：天津科技大学引进人才科研启动基金--图形图像分类技术的研究(No.20080436)

摘　　要：在色彩管理中,为了实现再现意图的自动选择,首先需要实现图形和图像的自动分类。本研究通过观察和分析大量不同类型原稿,得出图形与图像之间的本质差异在于颜色空间分布特性不同,并在此基础上,提炼出了具有显著区分能力的若干特征,并按照区分能力大小对其进行筛选。然后采用判别分析法,根据筛选得到的区分特征建立了图形图像判别函数。经检验,本方法的判别精度达到96.75%。In order to realize automatic choosing rendering intent in color management, the automatic classification for graph- ic and picture should be accomplished firstly. In this study, through analyzing the characteristics of a large number of ima- ges, it was confirmed that the essential difference between graphic and picture was the color spatial distribution characteris- tics. Based on that, several mathematical features were refined and selected in terms of their classification performance. With the selected features, discriminant analysis method was adopted to build up discriminant functions. Finally, the accuracy of the functions has been tested which is up to 96.75 %.

关 键 词：色彩管理 图像 图形 自动分类 空间特性 颜色特征 判别分析 

分 类 号：TS801.3[轻工技术与工程]