감성기반 이미지 검색을 위한 시각정보 요소별 감성공간 생성

Abstract

Nowadays. the fast advance of digital technology and the current web environment has been accelerating the field of information retrieval and processing. The Internet space using the Web is not strange any more to most people and they can obtain any information desired through the web. These changes have spawned a great deal of research aimed at enhancing service and convenience. So, many computer science researchers are committed to finding more useful and efficient methods to find appropriate results to meet users' needs. Among those, methods of my dissertation has been studied of Kansei(=Sensibility) information processing, the purpose is to develop human Kansei processing in computer.

Kansei is a very important factor for intelligent computing. Study for processing this Kansei is needed in many areas of computer science. The research of KANSEI information is a field aiming at processing and understanding "how human intelligence determines what is subjective information or ambiguous Kansei" and "how the information can be executed in a computer".

Among studies for realizing these aims, this paper focuses on the creation of Kansei Space for Kansei-based image retrieval. In other words, this study proposed the Kansei space for Kansei information processing concerning visual information. This dissertation has defined relation between visual information and Kansei as a fundamental research for Kansei based image retrieval. This dissertation has attempted various methods as statistical method, semantic differential method and others. This dissertation uses natural language that enables us to interpret intrinsic information behind Kansei and process high-level Kansei in terms of meaning through systematic lexical relations between Kansei.

As a result, this dissertation has created a Kansei space that is able to measure qualitative and quantitative values. Through the Kansei space, distances between Kansei and visual information can be measured. So I applied these Kansei spaces to images that are composed of elements of visual information and performed an application of retrieval based on Kansei. The paper demonstrated the availability of the created spaces through the evaluation of users’ satisfaction. Good satisfaction of users was achieved using the proposed methods of this paper. Through these procedures mentioned above, we were able to realize that a computer is able to process Kansei information about visual information.

This dissertation started with Kansei about color and expanded into shape, textures and patterns. That is a significant foundation for qualitative and quantitative data measurements of Kansei generated from visual data. It may not be an exaggeration to say that the paper is the first to research about the shape, texture and pattern. Until now, this paper considered the primary elements (color, shape, texture and pattern) among available visual information. Other factors that have more influence on human Kansei and expansion of the study range still needs to be done.