摘要
基于语义内容的图像检索是解决图像简单视觉特征和用户检索丰富语义之间存在的语义鸿沟的关键。其中情感语义是最高层的语义。本文首先介绍情感图像检索的一般框架,并引出情感图像检索的4个主要内容:图像感性特征的抽取、用户情感信息的计算机描述、情感用户模型的建立和用户模型的个性化;并对这4个主要内容的现有算法和研究进展进行归纳和总结;接着介绍3个典型的原型系统;最后从情感数据库、用户模型的评估和用户模型的计算3个方面阐明实现情感图像检索所面临的问题,并提出一些初步的解决思路。
Semantics image retrieval is the key issue to solve the semantic gap between users' complex semantics and visual features. Emotion is the most abstract semantic structure of images. This paper first introduces the general frame of emotion image retrieval and points out the four main research issues: to exact kansei features from images, to describe users' emotion information, to build affective user model and to individualize the model. Then some algorithms to solve these four issues are analyzed in detail. After that, three typical archetypal systems are presented. Finally, three critical problems, including emotion database, evaluation of user mode and computation of user model, faced in emotion image retrieval are explained, and some resolved strategies are presented elementarily.
出处
《电路与系统学报》
CSCD
北大核心
2005年第4期102-110,共9页
Journal of Circuits and Systems
基金
国家973资助基金课题-图像、语音、自然语言理解与知识挖掘(G1998030500)
校青年基金-感性信息处理及其在多媒体中的应用
国家自然科学基金-人工情感智能模型及其应用研究(60401004)
关键词
情感
图像检索
用户模型
个性化
emotion
image retrieval
user model
individualize