Eye-Perspective
Overview: Eye-perspective view management. Our view management utilizes synthesized EPR views for left and right eyes of a user to optimize label placement and improve legibility. Our method relies on four optimization criteria to control label placement: (1) background color uniformity, (2) lightness contrast with respect to the background color, (3) text legibility based on a frequency analysis using Gabor filters and (4) stereo vision uniformity ensuring that the scene background is visually the same for left and right eye. Colored pixels are areas that have been excluded as valid placement regions for labels based on the respective criterion. By combining all criteria valid placement regions are filtered. A vector field moves labels towards the closest valid region.
Abstract: Optical see-through (OST) head-mounted displays (HMDs) enable users to experience Augmented Reality (AR) support in the form of helpful real-world annotations. Unfortunately, the blend of the environment with virtual augmentations due to semitransparent OST displays often deteriorates the contrast and legibility of annotations. View management algorithms adapt the layout of annotations to improve legibility based on real-world information, typically captured by built-in HMD cameras. However, HMD camera views of the real world are distinctively different from the user's view through the OST display which decreases the final layout quality. We present a novel eye-perspective view management that utilizes synthesized high-fidelity renderings of the user’s view through the HMD to optimize annotation placement. Our method significantly improves over traditional camera-based view management in terms of annotation placement and legibility. Eye-perspective optimizations open up opportunities for further research on use cases relying on the user's true view through OST HMDs.
Acknowledgements: This work was supported by a grant from the Austrian Research Promotion Agency (grant no. 877104). Tobias Langlotz and Jonathan Sutton are supported by the Marsden Fund Council from Government funding (grant no. MFP-UOO2124)
Video:
