摘要
平面序列图像的立体化表述和任意视点生成是实现 3-D场景的关键技术。利用便携式摄像机沿着不同运动轨迹,连续对目标物体进行拍摄,可获得不同层次的包含有双目视差深度的平面视点序列图像。从获取的序列图像中选择不同的视点,即选择参考帧,利用与参考帧相邻的帧图像,估计参考帧位置的深度信息。采用自适应匹配算法,将各种层次的深度信息分配给图像的不同区域,形成具有不同深度层次的图像深度表示图,建立3-D观察场景。最后,通过实验对该算法进行了验证。
The three-dimensional (3-D) representation and view generation of 2-D image sequence is the key technique to the 3-D scene realization. The proposed approach consists in scanning a camcorder across several trajectories of the target object in order to generate a viewpoint-motion 2-D image sequence, in which various confidence levels depth maps of binocular image are included. The idea is to estimete depth at several prespecified locations, called 'reference frame', by using their neighbouring captured frames. We propose an adaptive matching algorithm, which assigns various confidence levels to different regions. Once the depth maps are computed at preselected viewpoint, the intensity and depth at these are used to reconstruct arbitrary views of a 3-D scene. Experimental results are presented to verify the approach.
出处
《中国民航学院学报》
2000年第4期36-43,共8页
Journal of Civil Aviation University of China
关键词
3-D场景
参考帧
自适应匹配算法
视频图像序列
three-dimensional scene
reference frame
estimate
depth
adaptive matching algorithm