An Empirical Evaluation of the Performance of Real-Time Illumination Approaches: Realistic Scenes in Augmented Reality

Abstract

Augmented, Virtual, and Mixed Reality (AR/VR/MR) systems have been developed in general, with many of these applications having accomplished significant results, rendering a virtual object in the appropriate illumination model of the real environment is still under investigation. The entertainment industry has presented an astounding outcome in several media form, albeit the rendering process has mostly been done offline. The physical scene contains the illumination information which can be sampled and then used to render the virtual objects in real-time for realistic scene. In this paper, we evaluate the accuracy of our previous and current developed systems that provide real-time dynamic illumination for coherent interactive augmented reality based on the virtual object’s appearance in association with the real world and related criteria. The system achieves that through three simultaneous aspects. (1) The first is to estimate the incident light angle in the real environment using a live-feed 360∘ camera instrumented on an AR device. (2) The second is to simulate the reflected light using two routes: (a) global cube map construction and (b) local sampling. (3) The third is to define the shading properties for the virtual object to depict the correct lighting assets and suitable shadowing imitation. Finally, the performance efficiency is examined in both routes of the system to reduce the general cost. Also, The results are evaluated through shadow observation and user study.