Blind Augmentation: Calibration-free Camera Distortion Model Estimation for Real-time Mixed-reality Consistency

This IEEE conference talk presents "Blind Augmentation," a novel approach for estimating camera distortion models without calibration to achieve real-time mixed-reality consistency. Learn how researchers from University College London, Birmingham City University, and University of Leeds tackle the challenge of matching augmented content with real camera footage affected by noise, motion blur, and depth of field. Discover their innovative method that instantly estimates parameters for these distortions, enabling the use of off-the-shelf real-time simulation methods in compositing. The presentation explains how modern computer vision techniques can remove these distortions from video streams, providing self-calibration that auto-tunes any black-box real-time method to deliver fast, high-fidelity augmentation consistency.