Computational Photography for Capturing Virtual Environments

用于导航、训练和计算机动画的基于图像的虚拟环境很有吸引力，但生成高质量和物理上合理的大型环境仍然超出了今天的技术。许多研究挑战仍然存在，包括在捕获期间光照变化的情况下使基于大图像的环境保持一致，并在不同的光照下显示它们。另一个主要挑战是赋予基于图像的环境一些交互物理行为，例如，对交互力的响应。基于图像的虚拟环境必须变得更加丰富，以提供这些新的功能，这些功能将使计算机动画行业、移动导航和安全、资产监控和基于模拟的培训等领域的新应用成为可能。提出的研究通过改变捕获过程，将计算直接纳入图像采集来解决这些挑战。捕获过程中的计算可以引导成像过程获取基于图像的环境所需的信息，例如，通过在线跟踪表面或通过调整图像曝光来捕获正确曝光的所有图像区域。捕获过程的可伸缩性将通过研究允许向现有环境添加新映像的增量方法来解决。增量方法还将有助于处理环境随时间变化的动态场景。这项研究将有极好的机会培养各领域的高素质人才，许多加拿大中小型企业准备接受这些人才。软件工程技能在今天的经济中是至关重要的，这项研究将在现实世界的传感器和相机，计算机视觉和移动并行计算在今天的图形处理单元（GPU）和CPU，以及人机交互方面创造软件专家，但它也将确保高素质的人才将获得基本的数学和物理背景，使他们脱颖而出，并确保他们的长期成功。

Image-based virtual environments for navigation, training and computer animation are attractive, but generating high-quality and physically plausible large environments is still beyond today's technology. Many research challenges remain, including making large image-based environments consistent despite changes in illumination during capture and displaying them afterwards under different illumination. Another major challenge is to give image-based environments some interactive physical behaviour, e.g., a response to interactive forces. Image-based virtual environments must become richer to provide these new capabilities which will enable novel applications in the computer animation industry, in mobile navigation and in security, in monitoring of assets and in simulation-based training. The proposed research addresses these challenges by changing the capture process to include computation directly into the image acquisition. Computation during capture can steer the imaging process to acquire the necessary information for the image-based environment, e.g., through on-line tracking of surfaces or by adjusting image exposure to capture all image areas correctly exposed. Scalability of the capture process will be addressed by investigating incremental methods which allow to add new images to an existing environment. Incremental methods will also help with dynamic scenarios where the environment evolves over time. The research will have excellent opportunities to train highly-qualified personnel in fields with many Canadian small and medium enterprises ready to take-up this personnel. Software engineering skills are critically important in today's economy and this research will create software experts in real-world sensors and cameras, in computer vision and in mobile parallel computing on today's graphics processing units (GPU) and on the CPU, as well as in human computer interaction but it will also ensure that highly-qualified personnel will acquire the fundamental mathematical and physical background to set them apart and ensure their long-term success.