摘要
图像情感语义的注释与检索起步不是很久,涉及了很多学科的综合知识,需要对心理学、计算机科学、生理学等各门学科的知识和前沿成果都有比较深入的了解,这个领域的研究充满了挑战和难度,同时其后续研究也存在着很大的可能性。情感语义是图像语义的最高层次,在图像情感语义注释和检索中起着很重要的作用。文中具体研究了底层特征提取中现有的一些常用方法,构建出图像的底层特征数据库。应用因子分析法对实验收集的用户情感数据库进行分析,构建出情感空间作为图像情感语义注释的基础。首次将LSSVM应用于图像情感语义注释上,实现了图像底层特征到高层情感语义的映射。然后通过相似度计算,在情感空间中完成图像的情感检索。实验结果取得了不错的用户满意度。
Image annotation and retrieval involves comprehensive knowledge of many disciplines, needs to be clear in the heart about psy- chology ,physiology,computer science and other subjects of knowledge and cutting-edge results. The research in this field is full of chal- lenge and difficulty ,but at the same time its follow-up studies also exists a lot of possibilities. Emotion semantic is the highest level of image semantics, it is extremely important in image emotion semantic annotation and retrieval. In this paper, study some existing common- ly used methods of underlying feature extraction, build up the image characteristics of the underlying database. Apply factor analysis meth- od m analyze the collection of user emotional database, and build emotional space as the basis of image emotional semantic annotation. LSSVM is applied to image semantic annotation for the first time,realizing the image characteristics of the underlying semantic mapping to the top. Then, through the calculation of similarity in the emotional space, complete the image retrieval. The experimental results have achieved good user satisfaction.
出处
《计算机技术与发展》
2015年第10期13-18,共6页
Computer Technology and Development
基金
上海市科学技术委员会资助项目(14590500500)
上海市自然科学基金(15ZR1415200)
上海高校青年教师培养资助计划(ZZSD13008)
关键词
“维量”思想
图像检索
情感语义注释
因子分析
LSSVM神经网络
"Dimensional" thinking
image retrieval
emotional semantic annotation
factor analysis
Least Squares Support Vector Ma-chine (LSSVM) neural network