다시점 비디오 부호화에서 영역기반 분석을 활용한 빠른 모드 예측 = (A) region analysis based fast mode decision for multi-view video coding

다국어 초록 (Multilingual Abstract)

A Multi-view video coding(MVC) which has been developed to improve the coding efficiency of multi-view video sequences, exploits inter-view correlation between cameras. However MVC has a problem of having higher complexity compared to the single view video coding due to the above property.
In this paper, we propose a region analysis based fast mode decision to reduce the computational time of the encoding process. The proposed scheme uses two methods. First, we proposed a method that does not use the interview prediction in cases when macroblocks are occluded. The reason is that there exists occluded area because of the location and angle of cameras in MVC. Second, we propose a region division based fast mode decision for multi-view video coding. In MVC, a fast moving region in the foreground tends to be predicted in disparity compensated prediction mode, while a stationary region in the background is likely to be predicted in motion compensated prediction mode. Thus, we can decide prediction mode of each region by exploiting disparity or motion information.
Experimental results shows that 15% ~ 22% of computational time can be saved compared to the JMVC 3.0 only with negligible coding efficiency losses.

목차 (Table of Contents)

• 1. 서론 1
• 2. 관련 연구 3
• 2.1 다시점 영상 부호화 과정 4
• 2.2 전역 변이 6
• 2.3 영역 변이 8
• 3. 영역기반 분석을 활용한 빠른 모드 예측 10
• 3.1 영역 분할을 활용한 빠른 모드 예측 10
• 3.1.1 영역 맵 생성 11
• 3.1.2 빠른 예측 모드 결정 19
• 3.2 제한적 시점 간 예측 구조 20
• 3.2.1 전역 변이를 이용한 영역 계산 방법 20
• 3.2.2 제한적 시점간 예측 기법 22
• 3.3 영역기반 분석을 활용한 빠른 모드 예측 알고리즘 25
• 4. 실험 및 성능 평가 28
• 4.1 실험 환경 28
• 4.2 성능 평가 29
• 5. 결론 및 향후 연구 과제 32
• Abstract 36