컬러와 시각정보 매칭을 이용한 감성기반 영상 검색시스템

Abstract

Today, with faster speed of the internet service and development of web-technology, a great amount of information and data with high added value is being distributed through the Internet network. There have been active movements to use information resources more effectively. Information retrieval means a series of behaviors that users retrieval information needed for decision making or collect specific information, retrieval information desired from the data saved in database, not information scattered and analyse them. In the beginning, it was text-based information retrieval, but its range has recently expanded to modelling, object recognition and even user interface. As the number of internet users has expanded, the amount of information has greatly increased and much information can be shared. It is difficult to find useful information in the internet network. So, to find information which satisfies our needs of information, we are developing new methods on the information retrieval and technology even though they are not efficient. Representative methods include XML, RDF and Semantic Web developed by W3C. And, as the era of the internet network came since late 1990s, it has been applied for various areas such as retrieval, computer vision, graphics and broadcasting. Though the vision management has such a short history, it has been come into spotlight in areas relating to video information based on visuality of human beings. As the amount of visual information has increased, we need an efficient retrieval method. Rather than low-level information such as color, texture and shape, we need more content-based information using expression of objects in images, background explanation and comments. Video retrieval, an area of technology which retrieve images according to the order of user inquiries, uses analysed information included in images. This study designs an image retrieval system with sensibility information by matching color information with visual information to retrieve images provided in the internetwork or pictures taken with a use of digital cameras. As content-based or comment-based retrieval provides images saved in DB according to judgement of managers, it is hard to obtain exact images desired by users. Therefore, this study aims at identifying same images through an analysis of color and intensity, classifies sensibility elements by matching visual elements according to textures, patterns, colors and types, and provide sensibility visual information by matching color values with visual elements that provide the results of image retrieval for users.