Moving-vector Estimation for 3-D Motion Analysis of Dynamic Image = 동영상의 3차원 분석을 위한 움직임요소 추정

국문 초록 (Abstract)

특징기반 동영상분석 방법에서 정확한 특징추출과 신뢰성있는 정합은 움직임 추정을 위한 중요한 요소이다. 본 연구에서는 특징기반 움직임 요소 추정방법과 스테레오 영상의 기하학적 구조를 연구하였다. 전체적으로 전처리 및 정합 과정으로 분리되며 전처리에서는 정교한 에지검출을 위하여 CL-map을 이용하였고 검출된 특징요소들을 정합을 위한 파라미터로 사용하였다. 스테레오 실험 영상을 이용하여 실험하였고 제안한 방법의 3차원 동작분석을 위한 타당성을 검증하였다.

다국어 초록 (Multilingual Abstract)

In the feature-based analysis of dynamic image, correct feature extraction and reliable matching are crucial factors for the motion estimation. In this paper, we present a feature-based approach to estimating the moving-vector of dynamic image and the geometric structure of binocular image sequences in a stationary environment. This scheme is a simple but robust feature-based stereo matching for the acquisition of 3-D moving vector. Feature-based methods are an effective paradigm for 3-D motion analysis. What is more such features as points, lines or even contours can be used as the input parameters in such methods.
The basic approachs here are divided into two portions: preprocessing and stereo matching. In the preprocessing·portions, we applied contour line map(CL-map) to improve the quality of the globally segmented region. Then these extracted moving-vectors can be used to set out the matching process on the binocular image sequences.
The proposed method has been tested on real images and, several experiments have been performed to show the feasibility that proposed approach can be used for 3-D motion analysis.