On 3D Content Manipulation: Simplification, Modification and Authentication

With the rapid development in the area of computer graphics and depth sensing technologies, various tools have emerged for 3D content creation such as RGB-D cameras (e.g., Microsoft Kinect, RealSense), LiDAR sensors, CAD tools etc. By employing these tools, creating and re ning the 3D data has become viable which in result has enabled numerous mixed-reality applications in the domain of healthcare, virtual training, collaborative visualization, vehicle automation and crime scene reconstruction. Although the 3D data nds application in diverse areas, deploying an immense 3D content across various platforms is still challenging. Besides, due to applicability in sensitive areas, scrutinizing the vulnerability of 3D data has also become crucial. Motivated by these challenges, this dissertation presents a set of novel approaches for 3D content manipulation mainly focusing on making it e ective across di erent platforms and reliable for various sensitive applications. Speci cally, I have considered three research tasks. The rst task focuses on 3D content simpli cation for multi-platform rendering that includes 1) designing a high- delity mesh simpli cation algorithm QEM4V R enabling mobile virtual reality (VR) and 2) a real-time curvature sensitive surface simpli cation CS3 using depth images. Among these, QEM4V R applies curvature based boundary preservation for generating a high- delity, low-poly version of 3D meshes in an o ine manner. While CS3 provides a real-time, curvature adaptive surface simpli cation from depth images for enabling mixed reality applications. We also studied the utilization of objective perceptual quality metrics to evaluate 3D mesh simpli cation algorithms that motivates us to formulate a Just Noticeable Di erence based subjective analysis. The second task is aimed at 3D content modi cation for virtual therapies that explores the applicability of 3D data in managing phantom pain as well as for a virtual enhancement providing the positive reinforcement during virtual therapy. Mixed reality-based framework developed for managing phantom pain (Mr.MAPP) creates a realistic and e ective illusion of a phantom limb in case of both upper and lower limb amputation. The system is evaluated by subject matter experts (SMEs) and has been enhanced for conducting the patient trial. The last task is 3D content authentication for secure usage by performing thorough studies on 3D data forensics for both long and short range sensors such as LiDAR and Microsoft Kinect respectively. Keeping the application of 3D data in self-driving cars, we also propose a framework, ALERT (Authentication, Localization, and Estimation of Risks and Threats), as a secure layer in the decision support system in the navigation control of vehicles and robots. Various experimental results demonstrate the e ectiveness of the proposed ALERT for ADAS (Advanced Driver Assistance System).