摘要
文中讨论了基于图像的VR环境图像生成中,未标定旋转图像序列的插补、整合及全景图的生成3 项关键技术问题,并给出插补、整合及全景图生成方法.所给方法有以下优点:插补和整合图像是源图像对应的摄像机在新视点下的视图,这保证了插补和整合图像的真实性;控制参数α变化,可得到所需视点方向的插补与整合图像,即可以控制视点方向;在此基础上开发的切点累积全景图生成技术克服了传统方法中技术缺陷,实用性强;此外,所给方法不需要事先标定摄像机.实验表明该方法处理简单。
In this paper, three key techniques are discussed in building VR visual field based on real images. They are interpolation, registration, and panorama of uncalibrated rotating image sequence. The merits of their implemented methods include the following, (1)The interpolated (registration) image accords with perspective geometry. This one ensures that the interpolated (registration) image is real; (2)When the coefficient( α ) is changed, the interpolated (registration) images corresponding to different views can be obtained. It means that the view direction can be controlled;(3)The tac\|point cumulation method of panoramic images overcomes some drawbacks in the previous methods; (4)The camera needn't be calibrated. The experiment indicates that the effect is very ideal.
出处
《计算机研究与发展》
EI
CSCD
北大核心
1999年第11期1349-1357,共9页
Journal of Computer Research and Development
基金
国家自然科学基金
关键词
图像插补
图像整合
图像生成
虚拟现实
计算机
matched matrix, image interpolation, image registration, panoramic images