首页> 外文期刊>ACM Transactions on Graphics >Unstructured Video-Based Rendering: Interactive Exploration of Casually Captured Videos
【24h】

Unstructured Video-Based Rendering: Interactive Exploration of Casually Captured Videos

机译:基于非结构化视频的渲染:随便捕获视频的交互式探索

获取原文
获取原文并翻译 | 示例
       

摘要

We present an algorithm designed for navigating around a performancernthat was filmed as a “casual” multi-view video collection:rnreal-world footage captured on hand held cameras by a few audiencernmembers. The objective is to easily navigate in 3D, generatingrna video-based rendering (VBR) of a performance filmedrnwith widely separated cameras. Casually filmed events are especiallyrnchallenging because they yield footage with complicatedrnbackgrounds and camera motion. Such challenging conditions precludernthe use of most algorithms that depend on correlation-basedrnstereo or 3D shape-from-silhouettes.rnOur algorithm builds on the concepts developed for the explorationrnof photo-collections of empty scenes. Interactive performerspecificrnview-interpolation is now possible through innovations inrninteractive rendering and offline-matting relating to i) modeling thernforeground subject as video-sprites on billboards, ii) modeling thernbackground geometry with adaptive view-dependent textures, andrniii) view interpolation that follows a performer. The billboards arernembedded in a simple but realistic reconstruction of the environment.rnThe reconstructed environment provides very effective visualrncues for spatial navigation as the user transitions between viewpoints.rnThe prototype is tested on footage from several challengingrnevents, and demonstrates the editorial utility of the whole systemrnand the particular value of our new inter-billboard optimization.
机译:我们提出了一种用于浏览表演的算法,该算法被录制为“休闲”多视图视频集合:由少数观众在手持摄像机上捕获的真实镜头。目的是轻松地在3D中导航,生成使用视频的渲染效果,该效果是通过视频拍摄的,由相距较远的摄像头拍摄的性能电影。随便拍摄的事件特别令人着迷,因为它们会产生具有复杂背景和摄像机运动的镜头。这种具有挑战性的条件使大多​​数算法都无法使用,这些算法依赖于基于相关性的立体或3D轮廓形状。我们的算法基于为空场景的探索照片集开发的概念。现在可以通过以下方式进行交互式表演者特定的视图插值:交互式渲染和脱机抠像,这些创新涉及以下方面:i)在广告牌上将地面主题建模为视频精灵; ii)使用自适应的视图相关纹理对背景几何进行建模; iii)跟随表演者的视图插值。将广告牌嵌入到简单但现实的环境中。随着用户在视点之间的转换,重建的环境为空间导航提供了非常有效的视觉效果。该原型在多个挑战性事件的镜头上进行了测试,展示了整个系统的编辑效用。新的广告牌间优化的特殊价值。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号