基于概念格的视觉单词约简方法  

A Reduction Method of Visual Words Based on Concept Lattice

作　　者：周亮亮[1] 

机构地区：[1]太原科技大学计算机科学与技术学院,太原030024

出　　处：《电脑开发与应用》2012年第9期15-17,共3页Computer Development & Applications

摘　　要：传统的视觉单词生成方式,仅通过无监督聚类方式获得,图像语义标注的精度和效率较低。采用概念格作为视觉单词约简工具,给出了一种新的视觉单词生成方法。首先,生成训练图像BOV模型的初始视觉词典,并将其归一化形成关于训练图像BOV模型的形式背景;其次构造BOV概念格,通过概念格的属性约简,实现对视觉单词的约简,从而生成最终的视觉单词。最后,通过实例表明了该方法是有效的。In order to get visual words, traditional method only through the unsupervised clustering method, image semantic annotation has low accuracy and efficiency. By use the concept lattice as visual word reduction tool, the paper presents a new visual words reduction method. First of all, generating BOV model visual dictionary of training images and normalization are into formal context, then construct the concept lattice, through attribute reduction realize visual word reduction. Finally, through the examples show that the method is effective.

关 键 词：图像语义标注 BOV 视觉单词 概念格属性约简 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]