基于感性的图像评估与检索  

KANSEI-BASED IMAGE EVALUATION AND RETRIEVAL

作　　者：王上飞[1] 陈恩红[1] 李金龙[1] 王煦法[1] 

机构地区：[1]中国科学技术大学计算机系,合肥230027

出　　处：《模式识别与人工智能》2001年第3期297-301,共5页Pattern Recognition and Artificial Intelligence

基　　金：国家973基金

摘　　要：感性信息处理的主要研究目标之一是让计算机能够模拟和识别用户的偏爱、喜好等主观信息,适应不同用户的不同需求,实现以人为本。它的理论和应用研究刚刚起步,是一个全新的研究课题,本文基于R.Plutchik的情感理论,建立了四维的情感空间,提出了人工情感模型,并将其应用于图像的感性评估和检索;采用径向基神经网络建立图像特征空间和情感空间的联系,记忆用户的感性信息,对图像进行感性评估;采用交互式遗传算法,将用户的情感结合到图像检索过程,实现图像的感性检索,取得了较好的实验结果。One of the main purposes of Kansei information processing is to imitate and recognize human psychology such as preference with computer, to adapt to different requirements of different users and to realize user-centered system. Both theory and application of this kind of research are still at the beginning. In this paper, based on the emotion theory of R. Plutchik, an artificial emotion model which is applied to Kansei-based image evaluation and retrieval has been proposed and a four-dimension affection space is established. A radial basis function neural network is used for learning and memorizing the user's emotion. Then each image is automatically evaluated with impression words in the four-dimension affection space. Interactive genetic algorithm is applied to realize image retrieval using human intuition and impression. Experimental results demonstrate the effectiveness of our approach.

关 键 词：图像评估 图像检索 感性 计算机 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]