由抖动图象序列建立自然环境三维真实感全景模型

由真实环境中的现场图象进行三维环境建模是目前国际上研究的热点问题。本文依据合理的运动模型，提出和实现了由包含抖动的摄像机运动下的图象序列建立３Ｄ环境全景模型的两步法。首先通过运动滤波和运动分解获得运动稳定的图象序列，然后采用无特征提取的时空纹理方向精确估计、深度边界确定和遮挡恢复算法，建立全局自然景物的真实感三维环境模型。提出了２种三维全景图象的表示方法，即非阵列方式深度分层区域表示和阵列方式的深度分层布景表示，可用于机器人全局定位的自然路标提取和真实环境虚拟再现的图象合成。该研究推广和结合了外极面图象的方法和全景图象的方法，放宽了对运动的要求，从而可使该种方法适用于室外颠簸的道路环境。和现有运动分层方法相比，避免了该类方法迭代过程中的局部最小化问题，并具有计算和存储效率高，适应性强，算法鲁棒性好的优点。

Building 3D Panoramic Model of Natural Scene from Fluctuated Image Sequences

The problem of virtualized reality, or view synthesis of real scenes, has received increasing attention in recent years. This paper presents a systematic approach to automatically construct the 3D natural scene from video sequences captured by a vibrated camera with only approximately known motion. A two stage method is presented for the 3D panoramic scene modeling which consists of (1) image stabilization by motion filtering and (2) depth estimation and depth boundary localization. The two stage method not only combines Zheng and Tsuji's panoramic image method with Baker's epipolar plane image analysis, resulting the so called panoramic epipolar plane image method, but also generalizes them to handle image sequence vibrations due to the uncontrollable fluctuation of the camera. The two stage method bypasses the correspondence problem and ill posed problem encountered in the general motion analysis, and avoids the local minimum problem in the spatial constraint based iteration method. This paper mainly describes the robust algorithms of spatio temporal orientation detection of spatio temporal textures, boundary localization of kinetic occlusion and depth layering of the panoramic scene. Two kinds of depth layered representations, depth layered region and depth layered setting, have been developed, which found a ground base for landmark selection of global localization and image synthesis of virtualized reality.