What to Distinguish and How? Opportunities and Challenges of Augmenting Multiple, Cluttered Objects in Complex Scenes for People with Low Vision

This paper investigates how Augmented Reality (AR) can assist people with low vision (PLV) in navigating complex, cluttered environments like busy kitchens or crowded streets. While previous AR systems for low vision focused on simple tasks or single-object identification, this research introduces 'AR distinction'—a method that visually differentiates multiple objects based on their importance level. The authors developed SceneGlance, a wearable AR system that uses RGB and depth data to identify objects and apply distinct visual overlays (e.g., solid colors for primary objects, outlines for secondary ones). The system was evaluated through a formative study, a controlled lab study with 12 participants in a mock-up kitchen, and a think-aloud study with 13 participants navigating outdoor routes.

The researchers found that AR distinction effectively guides user attention toward high-importance objects, such as safety hazards or task-critical tools. Participants reported that the system helped them build 'mental snapshots' of the scene and enabled hierarchical scanning, where they could prioritize information based on the visual cues provided. However, the study also identified a significant attention-recall tradeoff: while users were better at focusing on important items, the presence of multiple augmentations reduced their ability to recall other details of the scene. Furthermore, the researchers observed practical challenges, such as augmentations blending together in dense areas or failing to align correctly with objects in dynamic outdoor conditions.

As AR technology becomes more integrated into daily life, it offers immense potential for improving the independence of individuals with low vision. This work highlights that simply adding more visual information to a scene is not always beneficial; instead, designers must balance the need for guidance with the risk of cognitive and visual clutter. The findings provide a foundation for developing more sophisticated, context-aware AR systems that prioritize information based on user needs and environmental complexity.