Holostream: High-Accuracy, High-Speed 3D Range Video Encoding and Streaming
In recent decades, three-dimensional (3D) scanning techniques have improved in terms of speed, quality, and usability. These improvements have helped 3D scanning applications become increasingly popular within many different fields and industries, such as entertainment, security, manufacturing, and human computer interaction. However useful, real-time 3D scanning techniques generate large amounts of data, which may limit real-time storage or transmission of acquired 3D data. A platform which can enable high-quality, low-bandwidth 3D video communications could then be very useful within a broad range of applications, for example, telemedicine. The goal of telemedicine is to use telecommunications technology to remotely diagnose, monitor, and treat patients, offering: reduced health care costs, increased access to distant specialists, reduced health care inequity, improved patient comfort, and assistance in the early detection of adverse symptoms. Each evolution of communications technology has offered new advantages to the remote physicians to aid in their work. For instance, modern two-dimensional (2D) video communications can be used to allow remote physicians to both converse with and visually assess their patient's affliction, even if they are from a rural or underprivileged area. Although 2D photographs and videos work well for visual assessments, they lack depth and perspective which may be required to form an accurate diagnosis or to monitor a patient's affliction over time. Three-dimensional (3D) scanning systems, however, offer the potential to capture this information with great accuracy. Thus, if telemedicine technologies adopted 3D scanning systems, accurate measurements of afflicted areas could greatly influence the remote decision making of the physician. The goal of this dissertation research is to realize such a 3D communications platform that can transmit high-accuracy, high-resolution 3D data from one user to another over existing wireless network infrastructures. The first challenge in realizing an effective 3D communications platform is obtaining the high-quality 3D data to be delivered. In the context of telemedicine, for example, the accuracy, resolution, and field-of-view (FOV) of the 3D data could have significant impact on a physician's ability to perform analysis and make sound decisions. Structured light 3D scanners have the potential to provide both high-resolution scans with great accuracy; however, due in part to their calibration procedures, they are most effective at close distances (i.e., small FOV). In order to extend the measurement range of a structured light system, we proposed a novel camera calibration procedure which can more feasibly calibrate long-range vision systems. The procedure was able to accurately calibrate a camera (in-focus at a long range) at a near range (where the camera may be substantially out-of-focus) with only 0.2% difference in focal length versus a traditional, in-focus calibration. Our calibration procedure was then used to aid in the calibration of a long-range structured light 3D scanning system, extending its effective FOV. The second challenge in realizing an effective 3D communications platform is the transmission of 3D video: the information required to represent high-accuracy, high-resolution 3D data is quite large which makes it difficult to transmit quickly over standard wireless networks. Given this, we have proposed two techniques for encoding 3D data within a regular 2D image that are very resilient to lossy image compression; therefore, the encoded 2D image could then be substantially compressed using JPEG techniques. Our 3D data encoding procedures offered both very high compression ratios (i.e., small file sizes) and reconstruction accuracies (i.e., low error rates). For instance, one achieved a compression ratio of 935:1 when using JPEG 80, versus the OBJ file format, with a reconstruction error rate of 0.027%. With these accomplishments, transmitting high-quality 3D video data wirelessly to remote devices became possible. This dissertation research then developed the novel Holostream platform for high-quality 3D video recording, encoding, compression, decompression, visualization, and interaction. A demonstration system successfully delivered video-rate photorealistic 3D video content over standard wireless networks to mobile (e.g., iPhone, iPad) devices. Taking advantage of this dissertation research's two proposed 3D geometry encoding methods and mature video compression techniques, Holostream was able to deliver high-quality 3D video and color texture data to mobile devices using only 4.8-14 Mbps. Even under extremely poor network conditions, structurally representative 3D geometry and reasonably high-quality texture information could be delivered using only 0.48 Mbps. Finally, to emphasize the efficiency of this platform, this dissertation research also developed a fully mobile Holostream platform implementation that enabled mobile iPad devices to both send and receive 3D video content, acquired via a commercially available structured light scanner, in real-time across wireless networks. In summary, this dissertation research has (1) contributed methods for extending the effective FOV for high-accuracy, high-resolution 3D data capture; (2) contributed methods for the efficient and accurate encoding of high-resolution 3D range geometry data. (Abstract shortened by ProQuest.)
Allebach, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our