Acquisition of Large 3D Scenes - Dealing with Complexity and Incompleteness

在这项研究经费的续期申请中，我们建议继续我们的项目，该项目旨在通过3d扫描设备获取的真实场景的照片写实表现（建模，编辑，渲染），以实现在虚拟3d摄影和3d视频的道路上做开创性的工作。在资助的第一个阶段，我们开发了新的数据结构来处理实际扫描中的大量数据。我们实现了一个与项目合作伙伴共享的综合软件框架（XGRT）。有了这个框架，我们开发了算法来自动处理大型数据集，并以统计意义上的最佳方式重建表面。我们也取得了成功，并发表了一篇文章，在动画重建领域建立了一个新的研究方向，我们将继续我们的工作。我们展示了如何在真实世界的扫描中找到冗余，我们将在未来的工作中研究如何将其用于重建。

In this renewal application for a research grant, we propose to continue our project that aims at a photo-realistic representation (modelling, editing, rendering) of real-world scenes acquired by 3d-scanning devices to achieve the goal of doing the pioneering work on the way to virtual 3d photography and 3d video. In the first period of the grant we developed novel data structures to handle the large amount of data in real-world scans. We implemented a comprehensive software framework (XGRT) that we share with our project partners. With this framework we developed algorithms to process large datasets automatically and reconstruct surfaces in the best way in a statistical sense. We also had successes and a publication that established a new research direction in the field of animation reconstruction, where we will continue our work. We showed how redundancies in real-world scans can be found, and we will investigate in our future work how this can be utilized for reconstruction.