Foveated Encoding for Large High-Resolution Displays.

Area of Science:

• Computer Vision and Human-Computer Interaction
• Data Visualization and Display Technologies
• Image and Video Compression

Background:

• Collaborative analysis on high-resolution displays demands both high visual detail and low-latency data transfer, often creating a trade-off.
• Human visual perception is characterized by high acuity in the foveal region and reduced acuity in the periphery.

Purpose of the Study:

• To present a system that dynamically adapts image encoding quality for large displays.
• To reduce bandwidth requirements without sacrificing perceived visual detail during collaborative data exploration.

Main Methods:

• Gaze tracking of observers to identify the foveal region of interest.
• Dynamic adaptation of H.264 encoder quality parameters based on visual acuity fall-off.
• Implementation and testing using particle rendering and parallel coordinates visualizations.

Main Results:

• Substantial reduction in required bandwidth achieved through adaptive encoding.
• Minimal noticeable changes in visual quality, particularly within the foveated regions.
• Demonstrated high quality for collaborative analysis across distributed display walls.

Conclusions:

• Dynamically adapting encoding quality based on human visual perception is an effective strategy for bandwidth reduction.
• The proposed system enables efficient collaborative analysis of large scientific datasets on high-resolution displays.
• This approach balances visual fidelity and data transfer efficiency for remote collaboration.