3D Display Amblyopia Diagnosis via Eye Tracking

The patent implements personalized 3D rendering by detecting individual eye characteristics (pupil ellipses, iris patterns) and head pose for each viewer, then generating customized stereoscopic image pairs tailored to each person's anatomy and viewing geometry. This local customization resolves the contradiction by ensuring each viewer receives optimized 3D content while the system maintains accurate gaze tracking through dedicated computer vision algorithms.

The system performs preliminary face detection, eye landmark identification, and head pose estimation before generating 3D content. By pre-characterizing each viewer's eye geometry and viewing position, the system establishes accurate baseline measurements that enable both high-quality personalized rendering and precise gaze tracking throughout the viewing session.

The patent employs unsupervised deep learning models that automatically learn eye anatomy characteristics and gaze patterns from raw video data without requiring manual annotation or calibration sessions. The system self-calibrates by detecting pupil ellipses and iris patterns, then uses these learned models to perform both visual acuity assessment and drive 3D rendering, eliminating the need for complex external calibration equipment or expert intervention.

The deep learning system serves multiple functions simultaneously: it detects face landmarks, tracks eye gaze, assesses visual acuity, and generates personalized 3D content all through integrated computer vision pipelines. This multi-functionality reduces overall system complexity by consolidating what would otherwise require separate specialized subsystems into a unified deep learning framework.

The system performs preliminary detection of face landmarks, eye positions, and head pose, then caches these geometric parameters for use during 3D content generation. By pre-computing viewer-specific viewing geometry and storing it for reuse, the system minimizes real-time processing requirements while maintaining personalized rendering quality throughout the viewing session.

The patent implements foveated rendering that concentrates computational resources on the central foveal region where viewers fixate their gaze, using detected eye tracking data to allocate higher resolution processing to areas of visual interest while reducing processing in peripheral regions. This localized resource allocation maintains perceived quality while reducing overall computational burden.