High-Performance Image Acquisition and Processing for Stereoscopic Diagnostic Systems with the Application of...

Area of Science:

• Computer Vision
• Digital Cinematography
• Stereoscopic Imaging

Background:

• Three-Dimensional (3D) imaging is increasingly adopted in film, gaming, and AR, offering new creative possibilities.
• Ensuring viewer comfort and correcting 3D image errors during production is challenging, despite advancements in post-production.
• On-set detection of stereo vision errors, like depth budget or stereoscopic window violations, is crucial for efficient correction.

Purpose of the Study:

• To present a prototype diagnostic system for real-time calibration and depth analysis of stereoscopic cameras on film sets.
• To enable stereographers to identify and rectify 3D image errors, particularly those affecting viewing comfort, during the recording process.

Main Methods:

• Acquisition and synchronization of Full HD video streams from professional cameras via Serial Digital Interface (SDI).
• Estimation and analysis of disparity maps using uncalibrated rectification and GPU-accelerated algorithms.
• Real-time, non-invasive analysis of 3D depth without requiring intrinsic or extrinsic camera parameters.

Main Results:

• The developed system effectively analyzes 3D video depth, allowing immediate spotting of viewing comfort errors.
• GPU acceleration significantly enhances data processing efficiency for real-time analysis.
• The system achieves Full HD resolution processing at several frames per second, proving its on-set viability.

Conclusions:

• The diagnostic system provides an efficient and user-friendly solution for on-set stereoscopic camera calibration and 3D depth analysis.
• Real-time error detection and correction capabilities streamline the 3D filmmaking process, improving final image quality.
• The GPU-based approach makes complex 3D video analysis feasible in dynamic production environments.