Visual Complexity and Scene Recognition: How Low Can You Go?

Visual realism in Virtual Reality (VR) increases both immersion and development costs. Consequently, it is important to understand the cost-benefit trade-off of specific aspects of visual realism. Determining the extent of visual realism often leads to decisions on the level of visual complexity of a Virtual Environment (VE). In this paper, we investigate the impact of visual complexity on a user's spatial orientation through a user study. To do so, we created a VE containing parts of a real-world place. Participants were asked to map their location within the VE to the corresponding real-world location. They were provided by a pop-up map of the entire VE, on which they were required to choose one named location out of a predefined set of locations. We manipulated the VE's visual complexity by varying the visual elements used to provide cartographic information, namely a map overlay and 3D blocks as buildings. This results in four scene types: i) landscape contour, ii) landscape contour with 3D buildings, iii) landscape contour overlaid with a map, and iv) landscape contour with both 3D buildings and the map overlay. Each participant performed our location recall task for each scene type. Our findings provide empirical evidence that addition of each of these two visual elements individually improved spatial orientation, while their combination only adds a slight improvement.