Unconstrained Segue Navigation for an Immersive Virtual Reality Experience

Virtual reality (VR) is a rapidly growing field that has got innovative applications. Two main forms of VR content can be identified in the current applications --computer generated 3D models and adaptations of real world scenarios using digital imagery and video. However, real world adaptations are quite restrictive when it comes to interaction. These restrictions are due either to large storage needs, or to high computational needs, required to generate dynamic intermediate views. Thus, solutions through which real world simulation could be achieved in immersive and interactive VR environments are not available.

The research presented in this paper is intended to provide a high degree of interaction to users who engage in immersive VR experiences related to real world adaptations. The solution offered is a navigable grid map of spherical panoramas, which uses a three stage approach. Firstly, intermediate views among spheres are approximated using novel mechanisms. Secondly, an optimization strategy is introduced based on visual locality in which the areas with a higher probability of immediate interaction are given more prominence for quality during rendering. Thirdly, smooth segue transition is achieved through a machine learning backed gesture input system. The three approaches when combined together allow for an intuitive virtual experience while ensuring optimal resource utilization.