基于视频图像的虚拟环境建模技术  

The modeling technique of video-based virtual environment

作　　者：解凯[1] 郭恒业[1] 张田文[1] 

机构地区：[1]哈尔滨工业大学计算机科学技术学院

出　　处：《哈尔滨工业大学学报》2004年第4期471-473,共3页Journal of Harbin Institute of Technology

基　　金：国家自然科学基金资助项目(60075010);黑龙江教育厅科研基金资助项目(10531085).

摘　　要：针对基于平面投影变换的图像拼接方法存在的优化参数较多、易导致陷入局部极小和全景图像的生成速度缓慢的问题,提出一种优化参数较少的旋转运动模型,采用层次运动模型和傅里叶方法对帧间隔较大的图像进行对齐,使图像对齐工作量大大减少.该算法计算简单、快速,适用于手持相机拍摄的图像.The automated panoramic mosaicsing method based on planar projective transformation has more optimal coefficients so that it suffers from slow convergence and sometimes gets stuck in local minima, and it leads to generating slowly the panoramic image because the video has a lot of images. The paper presents the rotation motion model that has a fewer optimal coefficients. For a larger motion between adjacent frames, the algorithm uses Fourier transformations and hierarchical motion estimation framework to align them so that it can reduce the complexity of the algorithm greatly.

关 键 词：视频图像 虚拟环境 建模 平面投影变换 图像拼接 层次运动模型 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术] TP391.41[自动化与计算机技术—计算机科学与技术]