图像的情感语义研究进展

A Survey of Image Emotional Semantic Research

人对图像的感知和理解是主观性的。图像的情感语义研究的目的是希望计算机能够描述和表达人观察图像所引起的情感反应,使用带有主观感情色彩的语义表述图像,其结果可用于图像的情感评估、图像检索、美术和广告设计等方面,是一个崭新的且有很大挑战性前沿课题。本文从图像情感语义提取的结构框架出发,对图像的情感语义研究进展进行评述,包括了情感语义提取过程中的关键技术:图像特征提取、情感空间的选取与建立、建立图像特征到情感空间的映射等方面,并对国内外的研究现状进行跟踪研究。

Human perception and understanding of image are subjective. The motivation of image emotional semantic research is to endow the computer the ability for understanding image in emotional way as human being does. The results are to realize image subjective emotional semantic description, these results can be used in fields such as image emotional evaluation, image retrieval, design support, etc. It is a promising and challenging direction for research. In this paper, status quo of image emotional semantic research is reviewed. Typical system architectures are given, key techniques of the system are analyzed and discussed from three aspects: approaches to image feature extraction, schemes of human emotion description, approaches to build the mapping between image features and emotional space.