D-MO: Depth from Motion and Occlusion as a Visual Channel for Information Visualization

Paper Title

D-MO: Depth from Motion and Occlusion as a Visual Channel for Information Visualization

Publication Info

• Topic area: Development and evaluation of a novel visual channel for depth perception in information visualization.
• Keywords: Depth perception, motion parallax, occlusion, information visualization, visual channels, interaction design, user study, bubble charts, D-MO, expressiveness.

Background and Problem

• Problem / challenge: Depth as a visual channel is underutilized in information visualization due to its interference with size perception and ambiguity in encoding quantitative data.
• Significance: Expanding the set of usable visual channels can address the scarcity of channels for encoding ordered and quantitative data, enabling richer visualizations.
• Motivation and related work: Prior work in psychology and visualization has explored depth cues like motion parallax, occlusion, and stereokinetic effects. However, these approaches often require observer movement, specialized hardware, or introduce perspective distortions. There is a gap in leveraging motion and occlusion for depth perception without these limitations.

Solution

• Proposed approach: D-MO (Depth from Motion and Occlusion), a novel visual channel that combines relative motion and occlusion to convey depth without perspective distortion.
• Novelty:
  1. Introduces a depth channel based on motion and occlusion, avoiding perspective-related size distortions.
  2. Implements interaction and animation mechanisms for depth perception, including user-controlled projection angles.
  3. Demonstrates D-MO’s expressiveness and effectiveness through controlled experiments and integration into visualizations.
  4. Proposes a preliminary legend and interaction design for D-MO in real-world visualizations.
• Procedure and key techniques:
  1. D-MO alternates between left, center, and right projection angles, creating smooth animations.
  2. Interaction mechanisms allow users to manipulate the projection axis or pause the animation.
  3. Depth is encoded using occlusion and relative motion, with marks drawn in depth order.
  4. Controlled experiments evaluate D-MO’s accuracy, discriminability, and impact on size perception.

Results

• Concrete findings:
  • D-MO achieves a Stevens’ exponent of 1.04 (without interaction) and 0.86 (with interaction), indicating near-linear depth perception.
  • Log error for proportional judgments is 3.15 (without interaction) and 2.68 (with interaction), comparable to length-based channels.
  • D-MO supports up to five distinguishable depth levels with minimal overlap.
  • Size perception is slightly affected by D-MO, with a maximum distortion of ±4.8% in area.
• Advantage over baselines:
  • D-MO avoids the size distortion introduced by perspective-based depth channels (e.g., −25.5% to +29.2%).
  • Interaction significantly reduces order errors (1.3% with interaction vs. 10.8% without interaction).
• Experiments / evaluation:
  • Experiment 1: Depth perception study with 16 participants, showing low error rates and linearity in depth judgments.
  • Experiment 2: Size perception study with 16 participants, demonstrating minimal impact of D-MO on size perception.
  • Pre-study: Integration of D-MO into a bubble chart visualization, with 93% accuracy in answering questions.
• Limitations and future work:
  • D-MO’s effectiveness may vary across individuals, potentially requiring parameter customization.
  • Current implementation relies on occlusion and may be less effective with non-overlapping marks or alternative shapes.
  • Further studies are needed to refine interaction design, legend usability, and ghosting mechanisms for occluded marks.

Summary

D-MO introduces a novel depth visual channel based on motion and occlusion, enabling accurate and expressive depth perception without perspective distortion. Controlled experiments demonstrate its effectiveness in encoding ordered and quantitative data, with minimal interference with size perception. D-MO has been successfully integrated into visualizations like bubble charts and parallel coordinates, showing potential for broader applications. Future work will address individual variability, interaction refinements, and usability enhancements for occluded marks. This research expands the toolkit of visual channels for information visualization, offering new possibilities for complex data representation.