칼라특징을 이용한 내용기반 영상 검색 시스템 = A Content-Based Image Retrieval System Using Color Features

다국어 초록 (Multilingual Abstract)

In this paper, we are going to study a content-based image retrieval system using the color feature. The retrieval methods consists of histogram intersection and color pairs. The color of images is extracted after changing RGB(red, green, blue) color space to HSI(hue, saturation, intensity) that is the most compatible color for people's feeling. Hue and Intensity histogram thresholding for each color attribute are performed to split the chromatic regions and the achromatic regions. Region growing is used to solve oversegmentation problem. It is necessary to normalize the size of image for 160×128. Each image is divided into a smaller number of cells of 32×32 pixels. Color pairs are used to represent the query during retrieval.
The average Recall/Precision of histogram intersection is 0.66/0.69. The average Recall/Precision of color pairs is 0.47/0.60. The average Recall/Precision of proposed method is 0.76/0.79. The proposed method increases the retrieval effectiveness. The results demonstrate that the proposed method has led to substantial improvement in the performance of the system.

목차 (Table of Contents)

• Ⅰ. 서론
• Ⅱ. 내용기반 검색시스템의 개념
• 1. 내용기반 검색시스템의 구성
• 2. 텍스트 기반 검색의 문제점
• 3. 칼라를 이용한 검색 접근 방식
• 4. 칼라 공간
• 5. 히스토그램 인터섹션(Histogram Intersection)
• 6. 칼라쌍(color-pairs)
• Ⅲ. 제안된 내용기반 검색시스템
• 1. 정지화상 전처리기
• 2. 칼라 공간의 변환
• 3. 영역 분할
• 4. 색도, 명도에 의한 영역 분할
• 5. Oversegmentation 회복 단계
• 6. 히스토그램 생성기
• 7. 칼라쌍 생성기
• 8. 검색 및 매칭처리기
• Ⅳ. 실험결과 및 고찰
• 1. 실험환경
• 2. 실험 결과 및 분석
• 3. 고찰
• Ⅴ. 결론
• 참고문헌